Is there water on Earth's Moon ? updated 1998-05-14
This page is a summary of the claim I made on Foresight Exchange and some of the debate it generated.
------------------------------
Date: Fri, 6 Feb 1998 16:44:10 -0700 (MST)
From: "Patrik D'haeseleer" <patrik at cs.unm.edu>
Subject: fx-discuss: LunH2O news
I just approved the LunH2O claim. As a judge, I wanted to make sure first
that the claim creator agreed to a possibly extention of the deadline
in case of ambiguous results (considering the mapping resolution of the
Lunar Prospector is much coarser than required by the claim criteria).
However, rumors are already flying around regarding possible results
that may or may not already be in. Here's an article that appeared on
NASA Watch today:
6 February 1998: Has Lunar Prospector found Water on the Moon?
Yes, we have been hearing the same rumors all week too. They range
from huge amounts of water being found on the Moon to next to none
being found anywhere. We have also heard the rumors that some sort
of announcement is/was being considered for NASA HQ next week.
All we can say right now, with the rumor set's polarity flipping
back and forth on a daily basis (and our sources are all VERY good),
is that NASA and the Lunar Prospector team are keeping this all
very close to their chests and are working very hard to understand
their data.
Given the ill-prepared way that NASA PAO handled the press feeding
frenzy that surrounded the Mars meteorite announcement in August
1996, we hope that the scientists are allowed to do their analysis
in a careful, considered, peer-reviewed fashion and that all of
us in/near the press corps give them the elbow room they need to
figure all of this out.
Let's help them get it right. The implications are quite profound
if they have indeed found pre-positioned exploration provisions
on the nearest world to our own. Check out Luna NOW for the latest
on this mission.
BTW, NASA Watch (http://www.reston.com/nasa/watch.html) is IMHO definitely
THE place to keep an eye on anything involving NASA and space exploration,
including plenty of runors and sensitive internal data that NASA would
rather keep under wraps!
Hurry to get in now with a potentially very short-term claim!
Patrik
------------------------------
NASA Watch
http://www.reston.com/nasa/watch.html
------------------------------ Date: Wed, 8 Apr 1998 16:02:43 -0600 From: "David Blenkinsop" Subject: Re: fx-discuss: LunH20- ---------- > From: Patrik D'haeseleer > To: fx-discuss at ideosphere.com > Subject: Re: fx-discuss: LunH20 > Date: Tuesday, April 07, 1998 3:06 PM > > Erik Diehn writes: > > > >Given the continued ambiguity over this claim's resolution, I'd like to > >know how you'll judge if the deadline arrives and we haven't had a > >conclusive announcement about the amount of water on the Moon. As it stands > >now, if I'm correct, we *know* there's water, and we know that the upper > >ends of the projections would satisfy the claim -- but the lower end > >wouldn't. Would you extend the deadline for judging?Guys, as someone who holds some YES coupons,
I'm a little concerned about this call for extending the deadline, the comment about "the lower end" wouldn't satisfy the claim, also about Patrik's judges statement update of Mar 5th saying that the announcement of between 0.3% and 1% water ice is "not yet sufficient to judge the claim true"? Maybe I should have taken this judge's comment more seriously when I first read it, but doesn't an announcement of between 0.3% and 1% concentration over a very large area quite literally satisfy the claim?To see what I mean, go back up to the top of the claim's long description; the first thing it says is "Confirmation of "a lot"[*] of H2O on Luna announced by 1998-07-06". Now, this in itself is more then a little ambiguous, even given that "a lot" [*] is then defined as "at least 1 part in 200". Just *how* well does this result have to be confirmed?To see what is specified for "confirmation" of "a lot", check out the very next paragraph:
"This claim will be judged TRUE if, by 1998-07-06, NASA or associated scientists announce new[*] evidence of "a lot"[*] of H2O on the Moon, FALSE otherwise.
[*]"new": data collected after 1998-01-01.
[*] "a lot": a region over at least 50 square kilometers that averages
at least 1 part in 200 of H2O."
In other words, all that's needed is "announce new evidence" of "a region that averages at least 1 part in 200 of H2O". To rephrase that just slightly, we need an announcement of a large region that averages *0.5 percent* water content, I'm not misreading here, am I? Now, if you follow the URL that Patrik provides in the judge's statement, you find the following quote: "Binder and Feldman predict that water ice is confined to the polar regions and exists at only a 0.3 percent to 1 percent mixing ratio in combination with the Moon's rocky soil, or regolith". Looking at this as a statement of the mean or "average" amount expected, you would have to think that the most likely expected amount would be somewhere close to 0.65 percent, since that's the midpoint of the upper and lower bounds suggested. In fact, I'd go further and suggest that the scientists are almost certainly plotting their best guess as an estimated normal distribution, or probability density "bell curve", with 0.65 percent as the mean value under the highest part of the curve and the 0.3 percent and 1 percent numbers somewhat lower down on the curve on either side of this mean value.
> HOWEVER, I do expect the Explorer team to come up with more exact > results before the deadline. By now, they should be generating moon > maps of some of the more abundant minerals, plus ice. In fact, I will > send them some email to find out when they expect to release this > data. In the unlikely event that they have not published any new data > by the deadline, and that none is forthcoming, I would be inclined to > rule the claim false.
Ouch! I hate to say it, but if Patrik proceeds along this line, there is just no way that *any* reasonable amount of data will satisfy the claim, for years and years to come! My reasoning is that if the most probable mean value of the water concentration is *not* the deciding factor, then what number could *ever* be announced that would confirm the claim sufficiently well? For instance, someone could announce 0.55% as a probable lower bound, with 1% at the upper bound, as currently. At that point, some NO coupon holder could step in and say "ah, but they're not completely sure of that 0.55%, are they", i.e., you could always follow the normal probability curve down to 0.1% or whatever, if you want that much more certainty for that lower boundary! The thing that I'm really trying to get across here is that the claim *doesn't* specify absolute certainty about the confirmation of a water percentage, instead it only requires that "new evidence" be announced for the 1/2 percent concentration specified. In the spirit of what it means to have scientific evidence for such a matter, the claim *has* to be judged YES, in fact it really should have been judged so before now.
David Blenkinsop
------------------------------ End of fx-discuss-digest V2 #70 *******************************
------------------------------ Date: Thu, 9 Apr 1998 19:59:52 -0600 From: "David Blenkinsop" <blenl at sk.sympatico.ca> Subject: Re: fx-discuss: LunH20 On Wednesday, April 08, 1998, Daniel Rothman wrote: > > David Blenkinsop wrote: > > > > > > In other words, all that's needed is "announce new evidence" of "a region that > > averages at least 1 part in 200 of H2O". To rephrase that just slightly, we need an > > announcement of a large region that averages *0.5 percent* water content, I'm not > > misreading here, am I? Now, if you follow the URL that Patrik provides in the > > judge's statement, you find the following quote: "Binder and Feldman predict that > > water ice is confined to the polar regions and exists at only a 0.3 percent to 1 > > percent mixing ratio in combination with the Moon's rocky soil, or regolith". > > Looking at this as a statement of the mean or "average" amount expected, you would > > have to think that the most likely expected amount would be somewhere close to 0.65 > > percent, since that's the midpoint of the upper and lower bounds suggested. > > There is a distinction (real - not just semantic) between the most > probable estimate of a mean value, and the mean of several estimates. I'm not sure that I understand you here. I mean, the estimate of between 0.3 and 1 percent water is presumably an all-out best estimated average, over a very large area, by the people who should know these things. Now, are you saying that patchiness in the distribution will bring the maximum over some relatively small 50 square km area to something *less*? If this happens, the announced result directly implies that some neighboring patch would be even *more* concentrated to give the announced average in the end! Patchiness works in *favour* of the claim, but we shouldn't have to wait for an announcement of patchiness, since the announced average *already* fullfills the claim! > [David Blenkinsop wrote] > > In fact, > > I'd go further and suggest that the scientists are almost certainly plotting their > > best guess as an estimated normal distribution, or probability density "bell curve", > > with 0.65 percent as the mean value under the highest part of the curve and the 0.3 > > percent and 1 percent numbers somewhat lower down on the curve on either side of > > this mean value. > > Perhaps the judge should issue a precise statement of expected > probability. Such a statement might take the form, "if quantifiable, > new information must lead to an 85% probability that the H2O density in > a 50 km^2 region is at least 0.5% to form the basis of a yes judgement > to this claim." > > I would further suggest that the judge solicit such an opinion from the > researchers directly involved in this matter. This is what I'm now afraid has been on the judge's mind all along, accepting some particular, high confidence level for the result, a confidence level which is *not* written into the claim. Again the claim only requires that new evidence indicate the looked-for water fraction, no extra high level of confidence is required as proof. Surely, if it is true that the best estimate is a 0.65% fraction of H2O, then this is quite directly *more* than the 0.5% percent specified in the claim wording? The confidence level here is going to come out at *some* number corresponding to *higher* than 50/50 odds that the actual water fraction will exceed the claim's 1/2 percent -- the exact confidence level depends on the number of standard deviations from mean that the researchers decided to pick for their announced lower bound (0.3%) and higher bound (1%). Seems like *anything* higher than a 50% confidence, based on new evidence, ought to satisfy the claim wording, after all, 50% is the number that determines the preponderance of the evidence, whether the greater *weight* of the evidence is more on the YES side than on the NO, and so forth. David Blenkinsop <blenl at sk.sympatico.ca> ------------------------------ ... ------------------------------ Date: Fri, 10 Apr 1998 23:20:47 -0400 From: Daniel Rothman <drothman at geocities.com> Subject: Re: fx-discuss: LunH20 David Blenkinsop wrote: > On Wednesday, April 08, 1998, Daniel Rothman wrote: > > > > David Blenkinsop wrote: > > > > > > > > > > In other words, all that's needed is "announce new evidence" of "a region that > > > averages at least 1 part in 200 of H2O". To rephrase that just slightly, we need > an > > > announcement of a large region that averages *0.5 percent* water content, I'm not > > > misreading here, am I? Now, if you follow the URL that Patrik provides in the > > > judge's statement, you find the following quote: "Binder and Feldman predict that > > > water ice is confined to the polar regions and exists at only a 0.3 percent to 1 > > > percent mixing ratio in combination with the Moon's rocky soil, or regolith". > > > Looking at this as a statement of the mean or "average" amount expected, you would > > > have to think that the most likely expected amount would be somewhere close to > 0.65 > > > percent, since that's the midpoint of the upper and lower bounds suggested. > > > > There is a distinction (real - not just semantic) between the most > > probable estimate of a mean value, and the mean of several estimates. > > I'm not sure that I understand you here. I mean, the estimate of between 0.3 and 1 > percent water is presumably an all-out best estimated average, over a very large > area, by the people who should know these things. Yes - the all-out best estimate is between 0.3 and 1 percent over a wide area. Granted. This does _not_ say, nor should it be taken as saying, that the best estimate is 0.65%. You're taking a second-level estimate of an estimate, and calling it a first-level approximation. Combining estimates in this way is very dangerous, and only to be attempted by dedicated professionals using the most advanced safety equipment. If the current estimates were between 0.51 and 0.79, I don't think anyone would dispute the claim one bit. However, I don't think it's fair to fudge the numbers in the way you suggest - it's certainly not good statistics. Let me add, however, that the judge's discression is the final arbiter in this matter. > Now, are you saying that > patchiness in the distribution will bring the maximum over some relatively small 50 > square km area to something *less*? If this happens, the announced result directly > implies that some neighboring patch would be even *more* concentrated to give the > announced average in the end! Patchiness works in *favour* of the claim, but we > shouldn't have to wait for an announcement of patchiness, since the announced > average *already* fullfills the claim! > Nope, sorry. Not saying anything of the kind. I'm saying that the claim calls for a most probable value in a specific range (> 0.5%), and we don't have that. We have a most probable value that _may_ lie within that range. Or not. There is not currently sufficient evidence to make a valid statistical statement about whether or not the most probable value falls in the called-for range. > > [David Blenkinsop wrote] > > > In fact, > > > I'd go further and suggest that the scientists are almost certainly plotting their > > > best guess as an estimated normal distribution, or probability density "bell > curve", > > > with 0.65 percent as the mean value under the highest part of the curve and the > 0.3 > > > percent and 1 percent numbers somewhat lower down on the curve on either side of > > > this mean value. > > > > Perhaps the judge should issue a precise statement of expected > > probability. Such a statement might take the form, "if quantifiable, > > new information must lead to an 85% probability that the H2O density in > > a 50 km^2 region is at least 0.5% to form the basis of a yes judgement > > to this claim." > > > > I would further suggest that the judge solicit such an opinion from the > > researchers directly involved in this matter. > > This is what I'm now afraid has been on the judge's mind all along, accepting some > particular, high confidence level for the result, a confidence level which is *not* > written into the claim. Again the claim only requires that new evidence indicate the > looked-for water fraction, no extra high level of confidence is required as proof. Ah! here's the rub. The phrase "new evidence" is interpretable in many ways. You're currently using it synonymously with "new indications", meaning new data that does not directly disprove the idea. Unfortunately I, the judge, and the scientific community at large do not use the phrase in that manner. The meaning I read into that phrase is "new data directly supporting the conclusion", and by directly supporting, I mean in a very precise statistical way. The data from Lunar Prospector is precisely the sort that the two definitions disagree about. > Surely, if it is true that the best estimate is a 0.65% fraction of H2O, then > this is quite directly *more* than the 0.5% percent specified in the claim > wording? The confidence level here is going to come out at *some* number > corresponding to *higher* than 50/50 odds that the actual water fraction will > exceed the claim's 1/2 percent -- the exact confidence level depends on the > number of standard deviations from mean that the researchers decided to pick for > their announced lower bound (0.3%) and higher bound (1%). Seems like *anything* > higher than a 50% confidence, based on new evidence, ought to satisfy the claim > wording, after all, 50% is the number that determines the preponderance of the > evidence, whether the greater *weight* of the evidence is more on the YES side > than on the NO, and so forth. > > David Blenkinsop <blenl at sk.sympatico.ca> Not a bad argument if your statistics were valid. Unfortunately, it just isn't so. Let me try this way of expressing it: the mean of the end points of a confidence interval is not the same as the median of the underlying distribution. daniel rothman ------------------------------ ... ------------------------------ Date: Sat, 11 Apr 1998 16:21:46 -0600 From: "David Blenkinsop" <blenl at sk.sympatico.ca> Subject: Re: fx-discuss: LunH20 On Friday, April 10, 1998, Daniel Rothman wrote: > > > Yes - the all-out best estimate is between 0.3 and 1 percent over a wide area. Granted. > This does _not_ say, nor should it be taken as saying, that the best estimate is 0.65%. > You're taking a second-level estimate of an estimate, and calling it a first-level > approximation. Combining estimates in this way is very dangerous, and only to be > attempted by dedicated professionals using the most advanced safety equipment. > > If the current estimates were between 0.51 and 0.79, I don't think anyone would dispute > the claim one bit. However, I don't think it's fair to fudge the numbers in the way you > suggest - it's certainly not good statistics. Let me add, however, that the judge's > discression is the final arbiter in this matter. Hmm, at this point I think I have to admit to a mistake of terminology, one that seems to be deeply confusing the issue here. In my previous discussion, I used the term "confidence level" in a general sense of a predicted probability that one quantity would be larger than another, say that the water fraction would be greater than 0.3%. Thinking it over now, I realize that my math terminology is a bit rusty, and that statisticians use the word "confidence" for the very special situation where they want to know the uncertainty in one of the basic *parameters* of a distribution, like the uncertainty in the estimate of the mean, as such. This is a fine technical distinction or a reserved use for that particular word, and one that I didn't mean to bring up. **Throughout my previous discussion, I should have used the term "significance level", rather than "confidence level", apologies to all who were trying to relate this to the standard math usage! > [David Blenkinsop wrote:] > > Now, are you saying that > > patchiness in the distribution will bring the maximum over some relatively small 50 > > square km area to something *less*? If this happens, the announced result directly > > implies that some neighboring patch would be even *more* concentrated to give the > > announced average in the end! Patchiness works in *favour* of the claim, but we > > shouldn't have to wait for an announcement of patchiness, since the announced > > average *already* fullfills the claim! > > > > Nope, sorry. Not saying anything of the kind. I'm saying that the claim calls for a > most probable value in a specific range (> 0.5%), and we don't have that. We have a most > probable value that _may_ lie within that range. Or not. There is not currently > sufficient evidence to make a valid statistical statement about whether or not the most > probable value falls in the called-for range. OK, I managed to confuse the issue by substituting a wrong word, but aren't you managing to obfuscate in quite a serious way of your own? We've been given a definite range of values, for a water ratio - related measurement that is very likely to be normally distributed or distributed in a manner quite close to a normal distribution. Now you're saying that this announced range tells us *nothing* about a best, estimated or mean value of this distribution? Perhaps this is my fault, perhaps you are taking my mistaken usage of "confidence" too seriously? The thing I'd guess, here, if I'm not using any wrong words, is that the actual "confidence interval" for the mean value is going to quite narrow, given that they have a spacecraft in lunar orbit, a spacecraft that continuously beams back large numbers of measurements or data samples. According to my information, the estimated confidence interval is going to get narower as the square root of the total number of samples, so it just doesn't make sense, does it, that the confidence interval is going to be so large as to fill up the stated range between likely bounds, for the probable water fraction? Almost surely, they've got enough data to get the uncertainty in the basic parameters down quite low, it's the variability in the actual distribution that we should be discussing, that I've been trying to discuss, in fact, all along. In those terms, my basic idea is still true, I think -- the mean of the distribution is almost sure to be at the midpoint of the anounced lower and upper values. > > Not a bad argument if your statistics were valid. Unfortunately, it just isn't so. Let > me try this way of expressing it: the mean of the end points of a confidence interval is > not the same as the median of the underlying distribution. > > daniel rothman Say again? There may well be a theoretical point here, but in practical terms, well, just take any textbook example of estimating the mean value of a normal distribution, and you'll find just the opposite -- the point estimate of the mean is practically always taken as the center of the estimated confidence interval for that mean -- not that I really meant to bring up such a technicality in the first place! David Blenkinsop <blenl at sk.sympatico.ca> ------------------------------ End of fx-discuss-digest V2 #71 ******************************* Date: Wed, 22 Apr 1998 21:03:39 -0600 From: owner-fx-discuss-digest@@ideosphere.com (fx-discuss-digest) To: fx-discuss-digest at ideosphere.com Subject: fx-discuss-digest V2 #72 Reply-To: fx-discuss at ideosphere.com Sender: owner-fx-discuss-digest@@ideosphere.com Precedence: bulk fx-discuss-digest Wednesday, April 22 1998 Volume 02 : Number 072 ---------------------------------------------------------------------- Date: Sat, 11 Apr 1998 17:12:21 -0600 (MDT) From: "Patrik D'haeseleer" <patrik at cs.unm.edu> Subject: Re: fx-discuss: LunH20 David Blenkinsop writes: >We've been given a definite range of values, for a water ratio - >related measurement that is very likely to be normally distributed >or distributed in a manner quite close to a normal distribution. Now >you're saying that this announced range tells us *nothing* about a best, >estimated or mean value of this distribution? This is correct. We can *assume* a certain distribution and then figure out what the mean would be, but the fact is we have no clue what the real distribution is. Consider: - - Confidence interval for a normal distribution would commonly be given as 0.65 +/- 0.35. The fact that they chose to use 0.3 - 1.0 may be an indication that the distribution is skewed. - - Practically speaking, it may be that they have some approximately normal distribution for the strength of their signal but have to apply some nonlinear transformation to get the concentration of H2O. - - Normal distribution is almost certainly a bad fit, because the concentration of H2O can never be smaller than 0 (and the lower value, 0.3 %, is awfully close to zero). - - You could make some argument for LogNormal, but then the mean would be around 0.55 (sqrt(0.3 x 1.0)), which is too close to 0.5 to be comfortable. - - Since we're talking about a concentration, we should at least suggest a family of distributions that is only valid between 0% and 100%. >According to my information, the estimated confidence interval is going >to get narower as the square root of the total number of samples, so it >just doesn't make sense, does it, that the confidence interval is going >to be so large as to fill up the stated range between likely bounds, >for the probable water fraction? FYI, the 0.3% and 1.0% are not the absolute upper and lower bound to the concentration (that would be 0% and 100%), but most likely the confidence interval itself. I.e. the Explorer team has a certain degree of confidence (probably 95%) that the real concentration falls somewhere between these values. And yes, this confidence interval should go down over time. Question is whether then will release the new estimates before the deadline. If anybody wants to accost the Explorer team to find out what the current estimate of the concentration is, feel free to do so. However, I think they're probably being swamped with email, because I still haven't heard back from them. Patrik ------------------------------ Date: Sat, 11 Apr 98 16:38:59 PDT From: jim at mentat.com (Jim Gillogly) Subject: Re: fx-discuss: LunH20 Patrik says [regarding the distribution of lunar H20]: > - Confidence interval for a normal distribution would commonly be > given as 0.65 +/- 0.35. The fact that they chose to use 0.3 - 1.0 may > be an indication that the distribution is skewed. On their Web site they say "Uncertainties in the model mean even this estimated range may be off considerably." To me this means they aren't saying that in some places the concentration is 0.3%, in others it's 1.0%, and elsewhere it's somewhere in between, where the only question is whether the above-0.5% bits have enough area to satisfy the claim. I think it means that they really don't have a point estimate or even a confidence range for it. I don't claim certain knowledge or anything, of course -- it's not at all obvious how scientific results get translated into press releases. > If anybody wants to accost the Explorer team to find out what the current > estimate of the concentration is, feel free to do so. However, I think > they're probably being swamped with email, because I still haven't heard > back from them. I'm happy to have you harass them on our behalf. Jim Gillogly ------------------------------ Date: Sat, 11 Apr 1998 23:58:15 -0600 From: "David Blenkinsop" <blenl at sk.sympatico.ca> Subject: Re: fx-discuss: LunH20 Earlier, Patrik D'haeseleer wrote: > > David Blenkinsop writes: > >We've been given a definite range of values, for a water ratio - > >related measurement that is very likely to be normally distributed > >or distributed in a manner quite close to a normal distribution. Now > >you're saying that this announced range tells us *nothing* about a best, > >estimated or mean value of this distribution? > > This is correct. We can *assume* a certain distribution and then figure > out what the mean would be, but the fact is we have no clue what the > real distribution is. Consider: > > - Confidence interval for a normal distribution would commonly be > given as 0.65 +/- 0.35. The fact that they chose to use 0.3 - 1.0 may > be an indication that the distribution is skewed. > > - Practically speaking, it may be that they have some approximately normal > distribution for the strength of their signal but have to apply some > nonlinear transformation to get the concentration of H2O. While it may be worth speculating about skewness or nonlinearities, it is also worth noting that the normal distribution is well known for being a very good fit to many cases of randomness in continuous variables in the real world. One source that I've been perusing off and on for some time now is E.T. Jaynes' online textbook on Bayesian probability-as-logic theory at http://bayes.wustl.edu/Jaynes.book. If you download his Chapter 7, the chapter on the Gaussian distribution, you get quite a discussion there as to theoretical reasons why this distribution is so ubiquitous -- maybe this is an overly technical reference in some sense, but it's worth taking a look at, IMO. The presentation of Gauss's original idea of the normal distribution as a kind of "maximum likelihood" function seems particularly neat, once you begin to get the hang of it. The most basic idea here is that the Gaussian is generally the best representation of variability in continuous data under random or unknown influences, just the sort of thing that probabilities *should* be good at describing. In essence, it seemingly *does* take quite an odd, unusual or nonlinear situation to throw a variable "off normal" by much, so in a lot of cases, it would seem we could pretty much assume a normal distribution by default, unless there is some very special reason to think that the normal wouldn't apply. > > - Normal distribution is almost certainly a bad fit, because the > concentration of H2O can never be smaller than 0 (and the lower value, > 0.3 %, is awfully close to zero). Well, it's true enough that random quantities which have some natural, lower, limit, or upper limit, cannot be described by a normal distribution *exactly*, given that the pure, ideal or true Gaussian bell curve tails off asymptotically toward zero probability density, as the variable goes to infinity in either direction. For some sense of how much of a problem this is, practically speaking, try picking up a textbook or two, where practical examples of the normal distribution are discussed. For instance, from a printed textbook that I have, how about an example of a "normally distributed" data sample for thicknesses of iron plates coming out of a factory? The fact that the thickness of an iron plate can't be negative doesn't seem to be much of a concern -- apparently you just accept the Gaussian as a good approximation for values not too close to zero, and make some further adjustment or refinement if really needed as you get very close to that limiting number. As another, historical example, you could read the Jaynes chapter that I mentioned until you get to Jaynes' comments about Francis Galton's 1886 studies of "the heights of adult males in the population of England", pretty hard to make those heights negative! > > - You could make some argument for LogNormal, but then the mean would > be around 0.55 (sqrt(0.3 x 1.0)), which is too close to 0.5 to be > comfortable. > > - Since we're talking about a concentration, we should at least suggest a > family of distributions that is only valid between 0% and 100%. This would be analogous to looking at the thickness of iron plates in Log fashion, seemingly a strange or unhelpful maneuver, unless there's some special reason to assume some such nonlinearity? If such nonlinear ideas should apply in this case, I'm not sure what the reason would be, maybe it's better to assume the basic, "generated by linear randomness" normal curve as a good approximation? The logarithm thing is interesting, though, given that 0.3% is about "halfway to zero" from the center that I'm assuming of 0.65%; don't know what to suggest to really disprove this. > FYI, the 0.3% and 1.0% are not the absolute upper and lower bound to > the concentration (that would be 0% and 100%), but most likely the > confidence interval itself. I.e. the Explorer team has a certain degree > of confidence (probably 95%) that the real concentration falls somewhere > between these values. Maybe it's almost just as well that I made an "oops", earlier, substituting "confidence" for "significance", because "confidence" does seem to be in people's minds! To begin with, if I were announcing a number for a measured variable, I would *definitely* want to give people some idea of how far off *the actual number in question* might be from an estimated central value. In other words, I'd certainly want to put out a couple of numbers saying how low or how high the *true water fraction* might be expected to come out in the long run, where the true fraction is the actual amount of water divided by the amount of lunar regolith in question. Clearly, there is little chance that any estimated high and low numbers are "absolute bounds", they are only going to be "significance bounds", with my estimated distribution showing, say a 90% chance, or a 95% chance, (or whatever other significance number I may have chosen), that the actual fraction should be between those bounds. Note this is not the same as the usual technical meaning of "confidence interval", since the "confidence interval" is quite a restricted idea, it's an estimate of how dependably "right" you think your central, mean value is, but *then* you still have to take into account some kind of estimate for the *standard deviation* of the water fraction quantity that you are looking at! If I've got this straight, "confidence interval" is really meant for helping to decide how many data samples you need for dependable results, however, one's best model for the data is going to end up being a point estimate or central value, with effects of one's standard deviation estimate extending out to either side -- a bell curve, in other words, or minor variation thereof. Earlier, Jim Gillogly wrote: > > On their Web site they say "Uncertainties in the model mean even > this estimated range may be off considerably." To me this means > they aren't saying that in some places the concentration is 0.3%, > in others it's 1.0%, and elsewhere it's somewhere in between, where > the only question is whether the above-0.5% bits have enough area > to satisfy the claim. I think it means that they really don't have > a point estimate or even a confidence range for it. After all my comments on whether there is a "center" to the announced range of numbers, it's undoubtedly a good idea to ask whether *any* numbers so far are truly satisfactory or dependable in any sense. What Patrik should do here, IMO, is try to find out if there is truly any single, best estimate or point estimate of the overall water concentration, averaged out over some large volume of lunar regolith. If there *is* such a best estimate of overall water fraction, an estimate that people are basically satisfied with as truly meaningful, then *that's* what should be used to gauge the claim's fulfillment, at least until some better, perhaps more localized estimate comes along. In other words, it *shouldn't* be assumed that the lower end of every announced range of significance is the deciding factor -- the way this claim is worded, doing things *that* way is just as bad as saying that the high end of a range is the deciding number for the claim! Or so it would seem to me, anyway. David Blenkinsop <blenl at sk.sympatico.ca> ------------------------------ ...
------------------------------
Date: Wed, 06 May 1998 04:54:25 -0700
From: Jim Gillogly <jim at acm.org>
Subject: fx-discuss: LunH20: going bimodal
[Disclaimer: As of this writing I have no shares in
LunH2O, but I'm not promising to stay divested.]
I was just looking at the Lunar Prospector pages again,
and it looks like they may not get more specific for
another year, which would mean no tighter bounds on
the amount of water before the claim elapses:
http://lunar.arc.nasa.gov/science/results/lunarice/eureka.html
Further data analyses, as well as data from
another of Prospector's instruments, the gamma ray
spectrometer, will help mission scientists sort out the
precise distribution of lunar ice. The most informative
information is expected to be gleaned in just under a year,
when the spacecraft begins its extended mission and dips
down into a very low orbit of 10 kilometers above the
lunar surface.
On the other hand, I just looked at the claim statement
again, and I'm not confident that the intent of the claim
hasn't already been met. It defines "a lot" of water as
at least 0.5% H20 over at least 50 km^2. The NASA estimate
is anywhere between 0.3% H2O over 15000 km^2 to 1.0% H2O
over 70000 km^2. Their estimate of the absolute amount of
water is based on a depth of 0.5m, which is as deep as the
instruments measure; so the 10M-300M metric ton estimates
are presumably two different lower bounds depending on the
concentration and area, and their estimate of the number
of metric tons would be directly proportional to the area.
If we call that constant of proportionality k, then
0.3% * 15000 * k = 10M metric tons (with appropriate units
for k), and 1.0 * 70000 * k = 300M metric tons. This gives
k values of 200000 or 400000, which also suggests how many
significant figures we're working with here. Using these
k values, the LunH2O claim thus defines "a lot" of H2O as
being 0.5M - 1M metric tons.
The NASA URL above also says the final values may be off
by an order of magnitude in either case, which means the
actual number may be as low as 1M mt, coincidentally one
of the values suggested above; but this may be considered
a red herring, since they give best estimates of 10-300M mt.
The remaining issue is whether the claim "means" there has
to be a high concentration somewhere, or whether it's just
an indirect way to describe the total amount of H2O.
Executive summary: it seems likely we won't get more info
before the time is up, but there's at least one non-bozotic
interpretation of the claim that's already true.
- --
Jim Gillogly
15 Thrimidge S.R. 1998, 11:23
12.19.5.2.15, 2 Men 8 Uo, First Lord of Night
------------------------------
...
------------------------------
Date: Wed, 06 May 1998 08:35:43 -0500
From: Bill Schultz <pope at agnostic.org>
Subject: fx-discuss: Why Do So Many Claims Boil Down To What The Judge Thinks?
At 04:54 AM 5/6/98 -0700, Jim Gillogly wrote about LunH2O:
[Snip very long discussion of what the NASA findings
are versus what the claim specifies; including discussing
the current error estimates in the data which might
impact the eventual findings; plus the fact that the
claim due date is BEFORE the "final" data will be
announced.]
>The remaining issue is whether the claim "means" there has
>to be a high concentration somewhere, or whether it's just
>an indirect way to describe the total amount of H2O.
>
>Executive summary: it seems likely we won't get more info
>before the time is up, but there's at least one non-bozotic
>interpretation of the claim that's already true.
It all boils down to what the judge thinks, doesn't it?
=====
My greatest success on FX was on a claim where there was
literally no room for a judge's interpretation (HOST97 if
you are interested). The claim's judgment was defined in
terms of what number appeared on a given web page after a
certain occurrance. I figured out how to mathmatically
estimate the number, and when I became convinced of the
(correct!) answer, I poured almost all my assets into
what I viewed as a "sure thing" that some number in the
FX community seemed willing to bet against.
I surely thought UNIX would be another such claim. It
calls for judgment based on what number Dataquest publishes
for CY-2000 sales of UNIX operating systems. I've posted
the URL of the last measurement to this list. It shows
under 1.3 million per year as the current estimate, and a
"True" claim for UNIX is anything under 1.8 million per
year. I won't recapitulate the arguments which have been
beaten to death over whether or not Dataquest's numbers
are "right." I surely made the bulk of my investment in
the UNIX claim based on my belief that Dataquest's number
is Dataquest's number and the judgment for UNIX will be
based on Dataquest's number. Others in the FX community
seem to feel quite otherwise, and I guess if it turns out
that the judge for UNIX is among the UNIX aficionados, I
could stand to get really shafted (yet again).
You see, I have been "shafted" on a couple of other claims
where I thought the result was clearly in my favor, but
the judge found a way to rule that the claim went the
other way (based on a difference of opinion on what the
wording of the claim REALLY meant).
=====
All BS aside, most claims present a "risk factor" which is
entirely subjective: "What will the judge think?" In my
opinion, that is not a risk factor which exists in a REAL
futures exchange. Out there in the REAL world, judgment
day arrives ("triple witching" or whatever) and the claims
are disposed of, one way or the other, automatically
without any human interpretation of what the claim "means."
In the REAL world, futures contracts are very mechanical.
For those of you who have some interest in the reality of
the $vIF claim (real $ version of Idea Futures), this
matter of how to make claim judgment more mechanical would
seem to me to be the greatest stumbling block. Who wants
the money they receive (or worse, DON'T receive) to be
based on some subjective thought process of some judge?
in a real $ version of IF, each judgment with ANY kind of
subjective thought process in the decision would very
likely lead to litigation filed by the losers.
=====
Part of the problem seems to be in the process of deriving
the wording of the original claims. Jim's comments on the
mismatch between NASA's actual findings and the wording of
the LunH2O claim are right on point for demonstrating this
disparity. [I too own no LunH2O coupons, for what it is
worth.]
MY SUGGESTION: I'd re-work the "Standard Ambiguity Clause"
to include some kind of language to the effect that favor
will be shown to any interpretation which lends itself to
a more-exact mathmatical judgment. If applied to LunH2O,
that rule would result in the obvious (as Jim points out),
there is all but certainly enough lunar ice there to meet
the minimum number contemplated by the LunH2O claim, and
so (unless the time frame for the claim is extended), it
would seem to me that a "Yes" judgment should be a foregone
conclusion.
So, maybe LunH2O should be a test case for some post-facto
revision of the "Standard Ambiguity Clause" to ensure that
a "proper" judgment is called for?
- -- Bill Schultz JOIN THE AGNOSTIC CHURCH:
| bill at infidels.org pope at agnostic.org http://www.agnostic.org/
| bill at freethought.org http://www.infidels.org/org/singles/
| Internet Infidel: http://www.infidels.org/
------------------------------
Date: Wed, 06 May 1998 10:11:05 -0500
From: Don Bixler <dbixler_remove at uhc.com>
Subject: Re: fx-discuss: Why Do So Many Claims Boil Down To What The Judge Thinks?
Bill Schultz wrote:
>
> It all boils down to what the judge thinks, doesn't it?
>
There are a couple of issues here.
1) the (very real) difficulty of writing unambiguous claims
2) the prevalent notion that we can all obviously see what
author "meant".
(1) has been hashed to death and I think we are getting better
at it. (2) is still a real problem.
This came up about the time Ken implemented the judges statements.
(I don't remember which came first.) The majority opinion was
that "obvious intent" was the overriding concern when questions
came up. The thought was (as is explicitly stated in some judges
statements), "It's their claim, they ought to get the answer they
were looking for." (Paraphrased.)
I have a real problem with that.
The "obvious intent" is not always so obvious. Any cursory look
through the mail archives aught to make that "obvious". (Sorry :-)
As part of the analysis that should preceed any investment, you
should ask, "Is the claim well enough written that I can avoid
misinterpretations?" There are several claims that I would like
to invest in (or more in), but I have reservations as to the
wording. (The most notable is PicF. I have called twice for the
judges to give me some outlines as to the characteristics of a
picture phone, but have heard nothing.)
In my view, the wording should be the only defining factor for a
claim. In a few cases, this will lead to counter-intuitive results,
but I see this as the risk of trading in poorly written claims.
Of course, there do have to be some commonly understood principles.
For instance, this is the FUTURES Exchange. Requiring every claim
to state that "X will happen after date Y" is a bit of overkill.
There should be a presumption of "futureness". See the discussions
about closing URAN about six months ago.
> =====
>
> Part of the problem seems to be in the process of deriving
> the wording of the original claims. Jim's comments on the
> mismatch between NASA's actual findings and the wording of
> the LunH2O claim are right on point for demonstrating this
> disparity. [I too own no LunH2O coupons, for what it is
> worth.]
>
For LunH20, the author laid out specific conditions and they have
not yet been met. (I also hold a few NO coupons.) I find it
difficult to argue specific measurables in some cases, then claim
authors intent for others.
- --
Don Bixler e-mail: dbixler at uhc.com
bixler at wavefront.com
United HealthCare Corporation phone: 612-204-8808
"Outside of a dog, a book is man's best friend. Inside of a
dog, it's too dark to read."
Groucho Marx
------------------------------
Date: Wed, 06 May 1998 09:46:31 -0700
From: Jim Gillogly <jim at acm.org>
Subject: fx-discuss: Re: Why Do So Many Claims Boil Down To What The Judge Thinks?
Bill Schultz skribis:
[Snip very long discussion of what the NASA findings
Hey, sorry! I'll be terser this time!
> So, maybe LunH2O should be a test case for some post-facto
> revision of the "Standard Ambiguity Clause" to ensure that
> a "proper" judgment is called for?
I didn't mean my "analysis" of LunH2O to go so far as to say
what was proper. The <other> legitimate interpretation of
the claim statement is that it requires a high concentration
of H2O over a "big enough" area. This may have implications
for how easy it is to extract the water. Obviously setting
up a base in Tycho will be a lot harder if the highest
concentration is several parts per million than if it's 1%,
and it's a perfectly reasonable reading to interpret the claim
this way.
Second, as the author of the first draft of that ambiguity
clause, I didn't intend it to dictate the final answer, but just
provide the ground rule parameters for the rest of the judgement
criteria.
> In the REAL world, futures contracts are very mechanical.
As I said in an essay I wrote for the short-lived FX newspaper,
the biggest problem with judging claims is that it's not just
the outcomes that are difficult to predict, but rather the
very shape of the future. We <don't> know how NASA will report
the data. We <didn't> know that the Park Police would change
the way crowds were counted for the Million Man March. We
didn't have a <clue> that the gov't would hold off charging
the Unabomber after catching him for months, until after the
claim judgement date was closed. For the Tansu Ciller/PM claim,
the tie followed by a shared arrangement that turned out not to
work after all made the whole thing a crap shoot. This being
the case, for many very interesting claims, we're stuck with
judges. If it were like futures contracts, we'd have to
restrict the realm of claims to less interesting ones.
I think this kind of claim is still in the minority, though.
I'm hoping for a reasonable judgement on one claim (i.e. a
judge who reads it the way I do), but on most others I expect
the result to be unambiguous.
Oops, I promised to be shorter this time. Clip away!
- --
Jim Gillogly
15 Thrimidge S.R. 1998, 16:29
12.19.5.2.15, 2 Men 8 Uo, First Lord of Night
------------------------------
...
------------------------------
Date: Wed, 06 May 1998 20:29:11 +0100 (BST)
From: Stephen B Streater <stephen at surprise.demon.co.uk>
Subject: Re: fx-discuss: Why Do So Many Claims Boil Down To What The Judge Thinks?
On Wed 06 May, Don Bixler wrote:
> Bill Schultz wrote:
> > Part of the problem seems to be in the process of deriving
> > the wording of the original claims. Jim's comments on the
> > mismatch between NASA's actual findings and the wording of
> > the LunH2O claim are right on point for demonstrating this
> > disparity. [I too own no LunH2O coupons, for what it is
> > worth.]
> >
>
> For LunH20, the author laid out specific conditions and they have
> not yet been met. (I also hold a few NO coupons.) I find it
> difficult to argue specific measurables in some cases, then claim
> authors intent for others.
I am one of the more literal school. I think one of the problems
is that a lot of claim authors clearly want the claim to come
out true. So, if you really want to do what the author wanted,
they'll all be true! (eg abdt,FLT2,GoCh,HIVC,XLif2 etc).
In JaNP, which I own, I tried to put clear conditions down which
can be verified by anyone who wants to. I also chose a literal
minded judge (97). I don't expect much argument about the results -
but I'm prepared to be corrected on this.
- --
Stephen B Streater
------------------------------
Date: Wed, 6 May 1998 14:23:58 -0600 (MDT)
From: "Patrik D'haeseleer" <patrik at cs.unm.edu>
Subject: Re: fx-discuss: Why Do So Many Claims Boil Down To What The Judge Thinks?
Don Bixler writes:
>
>Bill Schultz wrote:
>>
>> It all boils down to what the judge thinks, doesn't it?
>>
>
>There are a couple of issues here.
>1) the (very real) difficulty of writing unambiguous claims
>2) the prevalent notion that we can all obviously see what
> author "meant".
>
>(1) has been hashed to death and I think we are getting better
>at it. (2) is still a real problem.
>
>This came up about the time Ken implemented the judges statements.
>(I don't remember which came first.) The majority opinion was
>that "obvious intent" was the overriding concern when questions
>came up. The thought was (as is explicitly stated in some judges
>statements), "It's their claim, they ought to get the answer they
>were looking for." (Paraphrased.)
>
>I have a real problem with that.
>
>The "obvious intent" is not always so obvious. Any cursory look
>through the mail archives aught to make that "obvious". (Sorry :-)
In the case of LunH2O, one could argue that the intent of the claim
has already been satisfied - if it weren't for the fact that the author
specifically intended the claim to measure also whether NASA would release
sufficient information soon enough. So now we have a two-part intent,
only one of which is fulfilled.
Personally, I would love to judge the claim YES asap. As the judge for
LunH2O, I'm in a dilemma, because not even the full intent of the author
is fulfilled, let alone the judging criteria that were explicitely written
down. As I've mentioned before, I had foreseen the possibility of such an
ambiguous situation, and tried to cover it by having the claim wording
changed. However, the author was very explicitely against leaving some
more leeway in the claim description, and I think it would be unfair
for me to go against that. I wound up having to accept the claim as
is because of time pressure (Prospector had started collecting data,
and could have announced initial results within days).
Either way, the decision to accept the preliminary Prospector results as
sufficient was a judgement call - that's what judges are for. Considering
the complete intent of the author, as well as the claim description,
I rejected the preliminary results. It's my decision, and I'll stick
to it, even though personally I would have preferred a different claim
wording and outcome. Changing my mind midstream now would be even more
unfair to the FX players.
Remember, folks: we still have two more months to go before the
deadline of this claim! I'm still expecting to hear more detained
results before that time. Yes, It's tru that Prospector will go to a
lower, higher resolution orbit after one year. But I believe that is no
longer considered part of the primary mission, and intended mainly to
milk Prospector as much as possible while it's still alive. It's almost
certain they will not wait that long to release further results.
Patrik
------------------------------
...
------------------------------
Date: Thu, 7 May 1998 00:37:10 -0400
From: tob at world.std.com (Tom Breton)
Subject: Re: fx-discuss: Why Do So Many Claims Boil Down To What The Judge Thinks?
Bill Schultz <pope at agnostic.org> writes:
> It all boils down to what the judge thinks, doesn't it?
[...]
> You see, I have been "shafted" on a couple of other claims
> where I thought the result was clearly in my favor, but
> the judge found a way to rule that the claim went the
> other way (based on a difference of opinion on what the
> wording of the claim REALLY meant).
>
> =====
[...]
> All BS aside, most claims present a "risk factor" which is
> entirely subjective: "What will the judge think?"
[...]
> In the REAL world, futures contracts are very mechanical.
>
> For those of you who have some interest in the reality of
> the $vIF claim (real $ version of Idea Futures), this
> matter of how to make claim judgment more mechanical would
> seem to me to be the greatest stumbling block. Who wants
> the money they receive (or worse, DON'T receive) to be
> based on some subjective thought process of some judge?
I agree that this is a major issue. I welcome any brainstorming on how
the judging process could be made less subjective.
We must accept that the original question will not be optimally worded.
It may not really ask an unambiguous question or its literal question
may prove to have little to do with the interesting question behind it
(the infamous URAN).
<approach>
One approach mite be to successively refine the meaning of a claim
*while trading*. That is, more-refined questions would be proposed
(which is normal) and the claim would be freely intertraded with the
more-refined questions (which is normal), but (here's the difference )
at an intermediate time a judge could decide that the original claim
best fitted one of the more-refined claims, or fit several more-refined
claims in some proportion, or of course fitted none. That's a smaller
decision than fully judging an ambiguous claim. Then shares in the
original claim would be replaced by shares of the chosen claim, or be
replaced proportionately. In this way a claim could be refined while
being traded.
For instance, an original claim mite say something like: "There's a
significant amount of water on the moon", and more-refined claims mite
say something like "There's at least 100,000 cubic meters of water on
the moon".
</approach>
Tom
------------------------------
------------------------------ Date: Tue, 7 Apr 1998 15:39:18 -0400 From: Erik Diehn <ediehn at mindspring.com> Subject: fx-discuss: LunH20 Given the continued ambiguity over this claim's resolution, I'd like to know how you'll judge if the deadline arrives and we haven't had a conclusive announcement about the amount of water on the Moon. As it stands now, if I'm correct, we *know* there's water, and we know that the upper ends of the projections would satisfy the claim -- but the lower end wouldn't. Would you extend the deadline for judging? Seems like a NO judgement wouldn't be entirely fair. After all, they may well have discovered the amount that we're after -- which would be before the deadline -- but haven't announced it publicly, and may not do so until after our deadline. Am I making sense? Erik Diehn ediehn at mindspring.com ------------------------------ Date: Tue, 7 Apr 1998 15:06:19 -0600 (MDT) From: patrik at cs.unm.edu (Patrik D'haeseleer) Subject: Re: fx-discuss: LunH20 Erik Diehn writes: > >Given the continued ambiguity over this claim's resolution, I'd like to >know how you'll judge if the deadline arrives and we haven't had a >conclusive announcement about the amount of water on the Moon. As it stands >now, if I'm correct, we *know* there's water, and we know that the upper >ends of the projections would satisfy the claim -- but the lower end >wouldn't. Would you extend the deadline for judging? Unfortunately, this is exactly the kind of ambiguous result I had in mind when I asked the claim creator to put a deadline extension clause into the claim. Considering the claim creator's reluctance to this, I will only extend the deadline if I have a reasonable assurance that the matter would be cleared up within a fairly short amount of time - say, if they anounce a new paper that would appear in Science a couple of weeks after the official deadline. HOWEVER, I do expect the Explorer team to come up with more exact results before the deadline. By now, they should be generating moon maps of some of the more abundant minerals, plus ice. In fact, I will send them some email to find out when they expect to release this data. In the unlikely event that they have not published any new data by the deadline, and that none is forthcoming, I would be inclined to rule the claim false. If it were mine, I would have written some more flexibility into the claim, but I'm not the creator... Comments, anyone? Patrik ------------------------------ ... ------------------------------ Date: Tue, 7 Apr 98 15:09:39 PDT From: jim at mentat.com (Jim Gillogly) Subject: Re: fx-discuss: LunH20 Patrik skribis: > Unfortunately, this is exactly the kind of ambiguous result I had in > mind when I asked the claim creator to put a deadline extension clause > into the claim. Considering the claim creator's reluctance to this, I > will only extend the deadline if I have a reasonable assurance that the > matter would be cleared up within a fairly short amount of time - say, > if they anounce a new paper that would appear in Science a couple of > weeks after the official deadline. ... > Comments, anyone? I feel your pain. On the one hand, it looks like the UNAB type of claim, which was really about whether they guy was going to be caught by the time the claim was written. (He was, but not indicted in time, and it was closed BANG on the dot with the wrong result. Am I still annoyed? You betcha!) In this case, it looks like the claim is about whether a whole bunch of H20 (where "whole bunch" is defined precisely) is found on the Moon. However, he goes on to say specifically that punters are guessing whether the results are released in a timely fashion: "... one is also betting on how quickly the results are made public." Since he's specifically interested in whether NASA's researchers are sitting on the data rather than releasing it, and says so in the claim, I'm not sure you're justified in extending it at all, no matter how soon they say they'll cough up more specific results. Seems to me that wording and intent are the same in this case. It would be nice if they make the discussion moot by releasing what they know. They certainly knew about the first round of results at least a week before the announcement, and kept it very close to the vest. [I'm not currently invested in LunH2O.] Jim Gillogly ------------------------------ ... ------------------------------ Date: Tue, 7 Apr 1998 17:09:31 -0600 (MDT) From: patrik at cs.unm.edu (Patrik D'haeseleer) Subject: Re: fx-discuss: LunH20 Jim Gillogly writes: > >Patrik skribis: >> Unfortunately, this is exactly the kind of ambiguous result I had in >> mind when I asked the claim creator to put a deadline extension clause >> into the claim. Considering the claim creator's reluctance to this, I >> will only extend the deadline if I have a reasonable assurance that the >> matter would be cleared up within a fairly short amount of time - say, >> if they anounce a new paper that would appear in Science a couple of >> weeks after the official deadline. > ... >> Comments, anyone? > >However, he goes on to say specifically that punters are guessing >whether the results are released in a timely fashion: "... one is also >betting on how quickly the results are made public." Since he's >specifically interested in whether NASA's researchers are sitting on >the data rather than releasing it, and says so in the claim, I'm not >sure you're justified in extending it at all, no matter how soon they >say they'll cough up more specific results. Seems to me that wording >and intent are the same in this case. Good point. However, considering the ambiguity in the current results I feel I could make an exception here. After all, the Explorer team has actually announced way *more* H20 than anybody predicted based on the Clementine data. In that sense, the intent of the claim is already satisfied. It's just that we don't know for sure whether there's at least one patch with a high enough concentration. Patrik ------------------------------ ... ------------------------------ Date: Wed, 8 Apr 1998 13:52:11 -0600 From: Cerebus The Aardvark <kreme at kreme.com> Subject: Re: fx-discuss: LunH20 At 15:09 -0700 04/07/1998, Jim Gillogly wrote: >However, he goes on to say specifically that punters are guessing >whether the results are released in a timely fashion: "... one is also >betting on how quickly the results are made public." Since he's >specifically interested in whether NASA's researchers are sitting on >the data rather than releasing it, and says so in the claim, I'm not >sure you're justified in extending it at all, no matter how soon they >say they'll cough up more specific results. Seems to me that wording >and intent are the same in this case. I would have to agree. You can't extend the deadline at all, it would violate the spirit of the claim. I do not own any shares in LH20 (at least I don't think I do) <Snail://80209/South%20Gaylord/1015/400> You think you're a MacAddict? My son is _named_ Macintosh <bell://303.722.2009> <http://www.kreme.com> ------------------------------ ...
------------------------------ Date: Wed, 08 Apr 1998 23:47:12 -0400 From: Daniel Rothman <dan.rothman at globalone.net> Subject: Re: fx-discuss: LunH20 This is a multi-part message in MIME format. - --------------F0ED28E3A8413458E5F23DE5 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit David Blenkinsop wrote: > > ---------- > > From: Patrik D'haeseleer <patrik at cs.unm.edu> > > To: fx-discuss at ideosphere.com > > Subject: Re: fx-discuss: LunH20 > > Date: Tuesday, April 07, 1998 3:06 PM > > > > Erik Diehn writes: > > > > > >Given the continued ambiguity over this claim's resolution, I'd like to > > >know how you'll judge if the deadline arrives and we haven't had a > > >conclusive announcement about the amount of water on the Moon. As it stands > > >now, if I'm correct, we *know* there's water, and we know that the upper > > >ends of the projections would satisfy the claim -- but the lower end > > >wouldn't. Would you extend the deadline for judging? > > Guys, as someone who holds some YES coupons, I'm a little concerned about this call > for extending the deadline, the comment about "the lower end" wouldn't satisfy the > claim, also about Patrik's judges statement update of Mar 5th saying that the > announcement of between 0.3% and 1% water ice is "not yet sufficient to judge the > claim true"? Maybe I should have taken this judge's comment more seriously when I > first read it, but doesn't an announcement of between 0.3% and 1% concentration over > a very large area quite literally satisfy the claim? > > To see what I mean, go back up to the top of the claim's long description; the first > thing it says is "Confirmation of "a lot"[*] of H2O on Luna announced by > 1998-07-06". Now, this in itself is more then a little ambiguous, even given that "a > lot" [*] is then defined as "at least 1 part in 200". Just *how* well does this > result have to be confirmed? e.g. an average value of 0.5% over a large area. Hence in some parts of the large area the concentration might be 0.0001%, in others 1% or even 5%. My reading of the text indicates tha this is the average value being discussed. > > To see what is specified for "confirmation" of "a lot", check out the very next > paragraph: > > "This claim will be judged TRUE if, by 1998-07-06, NASA or associated scientists > announce new[*] evidence of "a lot"[*] of H2O on the Moon, FALSE otherwise. > [*]"new": data collected after 1998-01-01. [*] "a lot": a region over at least 50 > square kilometers that averages at least 1 part in 200 of H2O." > > In other words, all that's needed is "announce new evidence" of "a region that > averages at least 1 part in 200 of H2O". To rephrase that just slightly, we need an > announcement of a large region that averages *0.5 percent* water content, I'm not > misreading here, am I? Now, if you follow the URL that Patrik provides in the > judge's statement, you find the following quote: "Binder and Feldman predict that > water ice is confined to the polar regions and exists at only a 0.3 percent to 1 > percent mixing ratio in combination with the Moon's rocky soil, or regolith". > Looking at this as a statement of the mean or "average" amount expected, you would > have to think that the most likely expected amount would be somewhere close to 0.65 > percent, since that's the midpoint of the upper and lower bounds suggested. There is a distinction (real - not just semantic) between the most probable estimate of a mean value, and the mean of several estimates. > In fact, > I'd go further and suggest that the scientists are almost certainly plotting their > best guess as an estimated normal distribution, or probability density "bell curve", > with 0.65 percent as the mean value under the highest part of the curve and the 0.3 > percent and 1 percent numbers somewhat lower down on the curve on either side of > this mean value. Perhaps the judge should issue a precise statement of expected probability. Such a statement might take the form, "if quantifiable, new information must lead to an 85% probability that the H2O density in a 50 km^2 region is at least 0.5% to form the basis of a yes judgement to this claim." I would further suggest that the judge solicit such an opinion from the researchers directly involved in this matter. > > > HOWEVER, I do expect the Explorer team to come up with more exact > > results before the deadline. By now, they should be generating moon > > maps of some of the more abundant minerals, plus ice. In fact, I will > > send them some email to find out when they expect to release this > > data. In the unlikely event that they have not published any new data > > by the deadline, and that none is forthcoming, I would be inclined to > > rule the claim false. > > Ouch! I hate to say it, but if Patrik proceeds along this line, there is just no way > that *any* reasonable amount of data will satisfy the claim, for years and years to > come! My reasoning is that if the most probable mean value of the water > concentration is *not* the deciding factor, then what number could *ever* be > announced that would confirm the claim sufficiently well? For instance, someone > could announce 0.55% as a probable lower bound, with 1% at the upper bound, as > currently. At that point, some NO coupon holder could step in and say "ah, but > they're not completely sure of that 0.55%, are they", i.e., you could always follow > the normal probability curve down to 0.1% or whatever, if you want that much more > certainty for that lower boundary! > > The thing that I'm really trying to get across here is that the claim *doesn't* > specify absolute certainty about the confirmation of a water percentage, instead it > only requires that "new evidence" be announced for the 1/2 percent concentration > specified. In the spirit of what it means to have scientific evidence for such a > matter, the claim *has* to be judged YES, in fact it really should have been judged > so before now. > > David Blenkinsop <blenl at sk.sympatico.ca> ...
more_lunh2o.txt
To: fx-discuss at ideosphere.com
From: David Cary <d.cary at ieee.org>
Subject: Re: fx-discuss: Why Do So Many Claims Boil Down To What The Judge Thinks?
Cc:
Bcc:
X-Attachments:
Yes, I wrote LunH20. What I really meant was ... well, I could tell you, but it's so much more entertaining to watch you all speculate on what I *really* meant. :-)
I suppose everyone who has ever written a claim really *thought* at the time that it was simple and unambiguous. I know I was surprised at how much speculation and interpretation and general hubbub my bit of prose has generated. How much "interpretation" can there be in deciding whether one number is less than some other number ? As it turns out, a surprisingly large amount of interpretation, not to mention statistical theory.
I'm sure that Patrik D'haeseleer thought this claim would be easy to judge. I know I did. (Sorry, Pat !)
It makes me wonder how often people choose the wrong interpretation for a ambiguous statement in a newspaper article, when neither the reporter nor the reader even realize the other interpretations even exist. (Or the results from a physics experiment.)
I appreciate the philosophical comments from Jim Gillogly. Yes, one of the more interesting things about the future is that we not only don't know the answers to our questions, but also don't even know the right questions to ask.
>Date: Thu, 9 Apr 1998 19:59:52 -0600
>From: "David Blenkinsop" <blenl at sk.sympatico.ca>
>Subject: Re: fx-discuss: LunH20
...
>This is what I'm now afraid has been on the judge's mind all along,
>accepting some particular, high confidence level for the result, a
>confidence level which is *not* written into the claim. Again the claim
>only requires that new evidence indicate the looked-for water fraction, no
>extra high level of confidence is required as proof.
...
>*anything*
>higher than a 50% confidence, based on new evidence, ought to satisfy the claim
>wording, after all, 50% is the number that determines the preponderance of the
>evidence
...
>David Blenkinsop <blenl at sk.sympatico.ca>
>Date: Fri, 10 Apr 1998 23:20:47 -0400
>From: Daniel Rothman <drothman at geocities.com>
>Subject: Re: fx-discuss: LunH20
...
>Yes - the all-out best estimate is between 0.3 and 1 percent over a wide area. Granted. This does _not_ say, nor should it be taken as saying, that the best estimate is 0.65%.
>You're taking a second-level estimate of an estimate, and calling it a first-level approximation. Combining estimates in this way is very dangerous, and only to be attempted by dedicated professionals using the most advanced safety equipment.
...
>Ah! here's the rub. The phrase "new evidence" is interpretable in many
>ways. You're currently using it synonymously with "new indications",
>meaning new data that does not directly disprove the idea. Unfortunately
>I, the judge, and the scientific community at large do not use the phrase
>in that manner. The meaning I read into that phrase is "new data directly
>supporting the conclusion", and by directly supporting, I mean in a very
>precise statistical way.
>Date: Sat, 11 Apr 1998 17:12:21 -0600 (MDT)
>From: "Patrik D'haeseleer" <patrik at cs.unm.edu>
>Subject: Re: fx-discuss: LunH20
...
>We can *assume* a certain distribution and then figure
>out what the mean would be, but the fact is we have no clue what the
>real distribution is. Consider:
>
>- - Confidence interval for a normal distribution would commonly be
> given as 0.65 +/- 0.35. The fact that they chose to use 0.3 - 1.0 may
> be an indication that the distribution is skewed.
...
>FYI, the 0.3% and 1.0% are not the absolute upper and lower bound to
>the concentration (that would be 0% and 100%), but most likely the
>confidence interval itself. I.e. the Explorer team has a certain degree
>of confidence (probably 95%) that the real concentration falls somewhere
>between these values.
>
>And yes, this confidence interval should go down over time. Question is
>whether then will release the new estimates before the deadline.
...
>Patrik
>Date: Sat, 11 Apr 1998 23:58:15 -0600
>From: "David Blenkinsop" <blenl at sk.sympatico.ca>
>Subject: Re: fx-discuss: LunH20
...
>One source that I've
>been perusing off and on for some time now is E.T. Jaynes' online textbook
>on Bayesian probability-as-logic theory at
>http://bayes.wustl.edu/Jaynes.book. If you download his Chapter 7, the
>chapter on the Gaussian distribution, you get quite a discussion there as
>to theoretical reasons why this distribution is so ubiquitous
...
>After all my comments on whether there is a "center" to the announced range of
>numbers, it's undoubtedly a good idea to ask whether *any* numbers so far
>are truly satisfactory or dependable in any sense. What Patrik should do
>here, IMO, is try to find out if there is truly any single, best estimate
>or point estimate of the overall water concentration, averaged out over
>some large volume of lunar regolith.
...
>David Blenkinsop <blenl at sk.sympatico.ca>
>Date: Wed, 06 May 1998 04:54:25 -0700
>From: Jim Gillogly <jim at acm.org>
>Subject: fx-discuss: LunH20: going bimodal
...
>The remaining issue is whether the claim "means" there has
>to be a high concentration somewhere, or whether it's just
>an indirect way to describe the total amount of H2O.
...
> Jim Gillogly
>Date: Wed, 6 May 1998 14:23:58 -0600 (MDT)
>From: "Patrik D'haeseleer" <patrik at cs.unm.edu>
>Subject: Re: fx-discuss: Why Do So Many Claims Boil Down To What The Judge Thinks?
...
>In the case of LunH2O, one could argue that the intent of the claim
>has already been satisfied - if it weren't for the fact that the author
>specifically intended the claim to measure also whether NASA would release
>sufficient information soon enough. So now we have a two-part intent,
>only one of which is fulfilled.
David Cary:
I'm obviously interpreting this paragraph contrary to what the author (Patrik D'haeseleer) intended :-).
Patrik seems to be saying that (a) part of the claim, if interpreted in a certain way, has already been satisfied, yet (b) NASA has not yet released "sufficient" information.
Given (b), how is it possible that Patrik knows (a) ? Perhaps helpful aliens informed Patrik of the true concentration of H2O on the moon ? :-)
>As the judge for
>LunH2O, I'm in a dilemma, because not even the full intent of the author
>is fulfilled, let alone the judging criteria that were explicitely written
>down.
...
>Remember, folks: we still have two more months to go before the
>deadline of this claim!
...
>It's almost
>certain they will not wait that long to release further results.
>
>
>Patrik
>Date: Wed, 06 May 1998 08:35:43 -0500
>From: Bill Schultz <pope at agnostic.org>
>Subject: fx-discuss: Why Do So Many Claims Boil Down To What The Judge Thinks?
...
>All BS aside, most claims present a "risk factor" which is
>entirely subjective: "What will the judge think?" In my
>opinion, that is not a risk factor which exists in a REAL
>futures exchange.
...
>in a real $ version of IF, each judgment with ANY kind of
>subjective thought process in the decision would very
>likely lead to litigation filed by the losers.
...
>- -- Bill Schultz
David Cary:
Yes, that is true, but in the "real world" some of the most important and critical decisions *are* made by subjective thought processes -- emergency medical decisions and legal court judgements come to mind. And in fact litigation is often filed relating to these decisions, just as you predicted. The fact that this *is* already acceptible in the real world tells me that perhaps it would be acceptable in a real $ version of IF.
>Date: Thu, 7 May 1998 00:37:10 -0400
>From: tob at world.std.com (Tom Breton)
>Subject: Re: fx-discuss: Why Do So Many Claims Boil Down To What The Judge Thinks?
>
>Bill Schultz <pope at agnostic.org> writes:
...
>> For those of you who have some interest in the reality of
>> the $vIF claim (real $ version of Idea Futures), this
>> matter of how to make claim judgment more mechanical would
>> seem to me to be the greatest stumbling block. Who wants
>> the money they receive (or worse, DON'T receive) to be
>> based on some subjective thought process of some judge?
>
>I agree that this is a major issue. I welcome any brainstorming on how
>the judging process could be made less subjective.
>
>We must accept that the original question will not be optimally worded.
>It may not really ask an unambiguous question or its literal question
>may prove to have little to do with the interesting question behind it
>(the infamous URAN).
...
[one clever approach snipped]
...
>
> Tom
David Cary:
Yes, I think that rather than trying to polish up each and every claim that comes through to eliminate as much ambiguity as possible, we should give some thought to trying to automatically cause claims to be unambiguous (or at least less ambiguous). I suspect that it is, in general, not possible to mechanically eliminate ambiguity. A proof (that my suspicion is true, or that it is false) would be interesting.
>Date: Wed, 06 May 1998 09:46:31 -0700
>From: Jim Gillogly <jim at acm.org>
>Subject: fx-discuss: Re: Why Do So Many Claims Boil Down To What The Judge Thinks?
...
>As I said in an essay I wrote for the short-lived FX newspaper,
>the biggest problem with judging claims is that it's not just
>the outcomes that are difficult to predict, but rather the
>very shape of the future. We <don't> know how NASA will report
>the data. We <didn't> know that the Park Police would change
>the way crowds were counted for the Million Man March.
...
>This being
>the case, for many very interesting claims, we're stuck with
>judges. If it were like futures contracts, we'd have to
>restrict the realm of claims to less interesting ones.
...
> Jim Gillogly
Good point. (I didn't know there was a newspaper -- are there any copies online ?)
----------------------------------------------------------------------
Date: Thu, 07 May 1998 11:42:25 +0200
From: Harald Tveit Alvestrand <Harald.Alvestrand at maxware.no>
Subject: Re: fx-discuss: Why Do So Many Claims Boil Down To What The Judge Thinks?
At 08:35 06.05.98 -0500, Bill Schultz wrote:
>At 04:54 AM 5/6/98 -0700, Jim Gillogly wrote about LunH2O:
>... Who wants
>the money they receive (or worse, DON'T receive) to be
>based on some subjective thought process of some judge?
Anyone who has ever brought a case before the court system?
>in a real $ version of IF, each judgment with ANY kind of
>subjective thought process in the decision would very
>likely lead to litigation filed by the losers.
....kicking the ball over to another set of subjective thought
processes......
kidding aside, an "appeals court" for FX judgment might not be a totally
bogus idea. But the implications are *not* trivial.
Harald A
- --
Harald Tveit Alvestrand, Maxware, Norway
Harald.Alvestrand at maxware.no
------------------------------
...
------------------------------
Date: Sat, 9 May 1998 22:42:44 -0600
From: "David Blenkinsop" <blenl at sk.sympatico.ca>
Subject: Re: fx-discuss: Why Do So Many Claims Boil Down To What The Judge Thinks?
On Wednesday, May 06, 1998, Don Bixler <dbixler_remove at uhc.com> wrote:
>
> >
>
> For LunH20, the author laid out specific conditions and they have
> not yet been met. (I also hold a few NO coupons.) I find it
> difficult to argue specific measurables in some cases, then claim
> authors intent for others.
Maybe I'm flogging a bit of a dead issue here, since the LunH20 judge, as well
as a few other people, too, seem to be "sort of" of this opinion, that the
March 5th Lunar Prospector announcement *doesn't* satisfy the claim. At least,
this seems to be the prevailing opinion in the sense that they would like
definitive evidence for a YES to be more persuasive than the announcement
provides. Where I still have great difficulty in agreeing is when I see this
sensibility phrased in just the way that Don has phrased it, that the claim
author's "specific conditions have not yet been met"? At the present moment, I
am completely divested of my previous YES holdings, however, let's just review
the key conditions that the claim's author stated, shall we, just for the
record? The criteria for the claim to be a YES by July 6th, 1998, are as
follows:
1) "NASA or associated scientists must announce new evidence of"
2) "1 part in 200 of H20"
3) "over at least 50 square kilometers"
Now, the March 5th announcement that we are talking about suggests that the
concentration is between "0.3% and 1% water ice" measured over such a large
area that one would definitely tend to think that it almost has to include at
least one region of 50 square kilometers, even allowing for large uncertainties
about the sparseness of the ice patches or deposits. Maybe it's a dumb
question, I don't know, but which part of the claim's criteria haven't been
met? Looking at point #1 above, the claim simply requires "evidence" for the
validity of points 2 and 3, so it's all really a judgement call, right, I mean,
it all hinges on what you regard as an acceptable standard of evidence?
Earlier, I made a suggestion, which I tried to back up with some ideas about
probability theory, that the *midpoint* of that 0.3% to 1% range would be a
good point of comparison for deciding where the balance of the evidence really
lies, given that the claim itself doesn't actually specify how to look at an
announced *range* of values. Now I can understand if most people find this to
be too easy a standard, or think that it would set a bad precedent in some way,
but still, the interpretation of this isn't really all that cut and dried, is
it?
David Blenkinsop <blenl at sk.sympatico.ca>
------------------------------
...
------------------------------
Date: Sun, 10 May 1998 21:18:45 -0500
From: Don Bixler <bixler at wavefront.com>
Subject: Re: fx-discuss: Why Do So Many Claims Boil Down To What The Judge Thinks?
At 10:42 PM 5/9/98 -0600, you wrote:
>
>
>On Wednesday, May 06, 1998, Don Bixler <dbixler_remove at uhc.com> wrote:
>
>>
>> >
>>
>> For LunH20, the author laid out specific conditions and they have
>> not yet been met. (I also hold a few NO coupons.) I find it
>> difficult to argue specific measurables in some cases, then claim
>> authors intent for others.
>
>Maybe I'm flogging a bit of a dead issue here, since the LunH20 judge, as
well
>as a few other people, too, seem to be "sort of" of this opinion, that the
>March 5th Lunar Prospector announcement *doesn't* satisfy the claim. At
least,
>this seems to be the prevailing opinion in the sense that they would like
>definitive evidence for a YES to be more persuasive than the announcement
>provides. Where I still have great difficulty in agreeing is when I see this
>sensibility phrased in just the way that Don has phrased it, that the claim
>author's "specific conditions have not yet been met"? At the present
moment, I
>am completely divested of my previous YES holdings, however, let's just
review
>the key conditions that the claim's author stated, shall we, just for the
>record? The criteria for the claim to be a YES by July 6th, 1998, are as
>follows:
>
>1) "NASA or associated scientists must announce new evidence of"
>
>2) "1 part in 200 of H20"
>
>3) "over at least 50 square kilometers"
>
If I were the judge (which I am not), I would not judge TRUE
until NASA issued a statement there the LOWER limit was higher
than the stated value. As I read the statements above, (1) and
(3) have been met. (2) hasn't.
Don Bixler
P.S. This will be my last statement on the subject, because, yes, it
is a bit like flogging a dead issue.
------------------------------
...
...
------------------------------
Date: Tue, 12 May 1998 22:12:32 -0600
From: "David Blenkinsop" <blenl at sk.sympatico.ca>
Subject: fx-discuss: LunH20, June PopSci News
Hi, was just reading the June 1998 "Popular Science" magazine's Science and
Technology Newsfront section, when I came across the following (apparently not
available on their website as yet):
' Prospector's spectrometer can spot a teaspoon of water in a cubic yard of
moon dirt, according to co-investigator William Feldman of Los Alamos National
Laboratory in New Mexico. "But it's finding 4 gallons per cubic yard spread out
over thousands of square miles in shadowed craters," he says.
Data collected during 1998 will precisely quantify the lunar water. Early
calculations show that the total could be between 11 million and 330 million
tons of ice. The larger number would fill a 4-square mile lake 35 feet deep.'
Let's see now, 4 gallons per cubic yard? My information says a U.S. gallon is
231 cubic inches, so 4 gallons per cubic yard comes to a fraction of 1.98 per
cent by volume. Now, I've always taken the LunH20 specification of "1 part in
200" to be a fraction by mass, so assuming a lunar regolith density 3 times
that of water, 1.98 per cent by volume becomes 0.66 per cent by mass! Haven't
I been saying all along that because 0.65 per cent was the midpoint of the
range announced on March 5th, that this ought to be considered the single most
likely, "point estimate" or expected amount for which we have evidence? Since
0.65% beats the "1 part in 200", or 0.5% specified in the claim, doesn't this
mean that the claim's specification *has* been met?
So far, the only definite thing that I've gotten on this from anyone else on FX
is the exceedingly gloomy idea that because the bottom end of the announced
range was down at 0.3%, this is absolutely what we must assume. In those terms
maybe the Popular Science interview means nothing to nobody, I dunno. What I
can't help but think is that if a real, live, scientist, William Feldman, finds
the 0.65% estimate to be the amount worth talking about, then maybe the claim
judge, among maybe a few others, have somehow managed to make a bad call on
this? Doesn't an assumption of more-or-less the *worst* case estimate actually
bias the claim toward the NO side, in the sense that it assumes a degree of
certainty which the claim wording doesn't require, and which even a scientist
wouldn't insist on, at least as a first, best estimate? Note that I presume
that not a lot of really definitive analysis has been completed since that
March 5th announcement -- I could be wrong, but I strongly suspect that Feldman
was just wording the estimate of 0.3% to 1% in the way that he found most
interesting and informative, by quoting the midpoint of the range, to be exact.
David Blenkinsop <blenl at sk.sympatico.ca>
------------------------------
...
------------------------------
Date: Sat, 16 May 1998 03:07:13 -0500
From: David Cary <d.cary at ieee.org>
Subject: Re: fx-discuss: Why Do So Many Claims Boil Down To What The Judge Thinks?
Yes, I wrote LunH20. What I really meant was ... well, I could tell you,
but it's so much more entertaining to watch you all speculate on what I
*really* meant. :-)
I suppose everyone who has ever written a claim really *thought* at the
time that it was simple and unambiguous. I know I was surprised at how much
speculation and interpretation and general hubbub my bit of prose has
generated. How much "interpretation" can there be in deciding whether one
number is less than some other number ? As it turns out, a surprisingly
large amount of interpretation, not to mention statistical theory.
I'm sure that Patrik D'haeseleer thought this claim would be easy to judge.
I know I did. (Sorry, Pat !)
It makes me wonder how often people choose the wrong interpretation for a
ambiguous statement in a newspaper article, when neither the reporter nor
the reader even realize the other interpretations even exist. (Or the
results from a physics experiment.)
I appreciate the philosophical comments from Jim Gillogly. Yes, one of the
more interesting things about the future is that we not only don't know the
answers to our questions, but also don't even know the right questions to
ask.
>Date: Thu, 9 Apr 1998 19:59:52 -0600
>From: "David Blenkinsop" <blenl at sk.sympatico.ca>
>Subject: Re: fx-discuss: LunH20
...
>This is what I'm now afraid has been on the judge's mind all along,
>accepting some particular, high confidence level for the result, a
>confidence level which is *not* written into the claim. Again the claim
>only requires that new evidence indicate the looked-for water fraction, no
>extra high level of confidence is required as proof.
...
>*anything*
>higher than a 50% confidence, based on new evidence, ought to satisfy the
>claim
>wording, after all, 50% is the number that determines the preponderance of the
>evidence
...
>David Blenkinsop <blenl at sk.sympatico.ca>
>Date: Fri, 10 Apr 1998 23:20:47 -0400
>From: Daniel Rothman <drothman at geocities.com>
>Subject: Re: fx-discuss: LunH20
...
>Yes - the all-out best estimate is between 0.3 and 1 percent over a wide
>area. Granted. This does _not_ say, nor should it be taken as saying,
>that the best estimate is 0.65%.
>You're taking a second-level estimate of an estimate, and calling it a
>first-level approximation. Combining estimates in this way is very
>dangerous, and only to be attempted by dedicated professionals using the
>most advanced safety equipment.
...
>Ah! here's the rub. The phrase "new evidence" is interpretable in many
>ways. You're currently using it synonymously with "new indications",
>meaning new data that does not directly disprove the idea. Unfortunately
>I, the judge, and the scientific community at large do not use the phrase
>in that manner. The meaning I read into that phrase is "new data directly
>supporting the conclusion", and by directly supporting, I mean in a very
>precise statistical way.
>Date: Sat, 11 Apr 1998 17:12:21 -0600 (MDT)
>From: "Patrik D'haeseleer" <patrik at cs.unm.edu>
>Subject: Re: fx-discuss: LunH20
...
>We can *assume* a certain distribution and then figure
>out what the mean would be, but the fact is we have no clue what the
>real distribution is. Consider:
>
>- - Confidence interval for a normal distribution would commonly be
> given as 0.65 +/- 0.35. The fact that they chose to use 0.3 - 1.0 may
> be an indication that the distribution is skewed.
...
>FYI, the 0.3% and 1.0% are not the absolute upper and lower bound to
>the concentration (that would be 0% and 100%), but most likely the
>confidence interval itself. I.e. the Explorer team has a certain degree
>of confidence (probably 95%) that the real concentration falls somewhere
>between these values.
>
>And yes, this confidence interval should go down over time. Question is
>whether then will release the new estimates before the deadline.
...
>Patrik
>Date: Sat, 11 Apr 1998 23:58:15 -0600
>From: "David Blenkinsop" <blenl at sk.sympatico.ca>
>Subject: Re: fx-discuss: LunH20
...
>One source that I've
>been perusing off and on for some time now is E.T. Jaynes' online textbook
>on Bayesian probability-as-logic theory at
>http://bayes.wustl.edu/Jaynes.book. If you download his Chapter 7, the
>chapter on the Gaussian distribution, you get quite a discussion there as
>to theoretical reasons why this distribution is so ubiquitous
...
>After all my comments on whether there is a "center" to the announced range of
>numbers, it's undoubtedly a good idea to ask whether *any* numbers so far
>are truly satisfactory or dependable in any sense. What Patrik should do
>here, IMO, is try to find out if there is truly any single, best estimate
>or point estimate of the overall water concentration, averaged out over
>some large volume of lunar regolith.
...
>David Blenkinsop <blenl at sk.sympatico.ca>
>Date: Wed, 06 May 1998 04:54:25 -0700
>From: Jim Gillogly <jim at acm.org>
>Subject: fx-discuss: LunH20: going bimodal
...
>The remaining issue is whether the claim "means" there has
>to be a high concentration somewhere, or whether it's just
>an indirect way to describe the total amount of H2O.
...
> Jim Gillogly
>Date: Wed, 6 May 1998 14:23:58 -0600 (MDT)
>From: "Patrik D'haeseleer" <patrik at cs.unm.edu>
>Subject: Re: fx-discuss: Why Do So Many Claims Boil Down To What The Judge
>Thinks?
...
>In the case of LunH2O, one could argue that the intent of the claim
>has already been satisfied - if it weren't for the fact that the author
>specifically intended the claim to measure also whether NASA would release
>sufficient information soon enough. So now we have a two-part intent,
>only one of which is fulfilled.
David Cary:
I'm obviously interpreting this paragraph contrary to what the author
(Patrik D'haeseleer) intended :-).
Patrik seems to be saying that (a) part of the claim, if interpreted in a
certain way, has already been satisfied, yet (b) NASA has not yet released
"sufficient" information.
Given (b), how is it possible that Patrik knows (a) ? Perhaps helpful
aliens informed Patrik of the true concentration of H2O on the moon ? :-)
>As the judge for
>LunH2O, I'm in a dilemma, because not even the full intent of the author
>is fulfilled, let alone the judging criteria that were explicitely written
>down.
...
>Remember, folks: we still have two more months to go before the
>deadline of this claim!
...
>It's almost
>certain they will not wait that long to release further results.
>
>
>Patrik
>Date: Wed, 06 May 1998 08:35:43 -0500
>From: Bill Schultz <pope at agnostic.org>
>Subject: fx-discuss: Why Do So Many Claims Boil Down To What The Judge Thinks?
...
>All BS aside, most claims present a "risk factor" which is
>entirely subjective: "What will the judge think?" In my
>opinion, that is not a risk factor which exists in a REAL
>futures exchange.
...
>in a real $ version of IF, each judgment with ANY kind of
>subjective thought process in the decision would very
>likely lead to litigation filed by the losers.
...
>- -- Bill Schultz
David Cary:
Yes, that is true, but in the "real world" some of the most important and
critical decisions *are* made by subjective thought processes -- emergency
medical decisions and legal court judgements come to mind. And in fact
litigation is often filed relating to these decisions, just as you
predicted. The fact that this *is* already acceptible in the real world
tells me that perhaps it would be acceptable in a real $ version of IF.
>Date: Thu, 7 May 1998 00:37:10 -0400
>From: tob at world.std.com (Tom Breton)
>Subject: Re: fx-discuss: Why Do So Many Claims Boil Down To What The Judge
>Thinks?
>
>Bill Schultz <pope at agnostic.org> writes:
...
>> For those of you who have some interest in the reality of
>> the $vIF claim (real $ version of Idea Futures), this
>> matter of how to make claim judgment more mechanical would
>> seem to me to be the greatest stumbling block. Who wants
>> the money they receive (or worse, DON'T receive) to be
>> based on some subjective thought process of some judge?
>
>I agree that this is a major issue. I welcome any brainstorming on how
>the judging process could be made less subjective.
>
>We must accept that the original question will not be optimally worded.
>It may not really ask an unambiguous question or its literal question
>may prove to have little to do with the interesting question behind it
>(the infamous URAN).
...
[one clever approach snipped]
...
>
> Tom
David Cary:
Yes, I think that rather than trying to polish up each and every claim that
comes through to eliminate as much ambiguity as possible, we should give
some thought to trying to automatically cause claims to be unambiguous (or
at least less ambiguous). I suspect that it is, in general, not possible to
mechanically eliminate ambiguity. A proof (that my suspicion is true, or
that it is false) would be interesting.
>Date: Wed, 06 May 1998 09:46:31 -0700
>From: Jim Gillogly <jim at acm.org>
>Subject: fx-discuss: Re: Why Do So Many Claims Boil Down To What The Judge
>Thinks?
...
>As I said in an essay I wrote for the short-lived FX newspaper,
>the biggest problem with judging claims is that it's not just
>the outcomes that are difficult to predict, but rather the
>very shape of the future. We <don't> know how NASA will report
>the data. We <didn't> know that the Park Police would change
>the way crowds were counted for the Million Man March.
...
>This being
>the case, for many very interesting claims, we're stuck with
>judges. If it were like futures contracts, we'd have to
>restrict the realm of claims to less interesting ones.
...
> Jim Gillogly
Good point. (I didn't know there was a newspaper -- are there any copies
online ?)
- --
+ David Cary "mailto:d.cary@ieee.org" "http://www.rdrop.com/~cary/"
| Future Tech, Unknowns, PCMCIA, digital hologram, <*> O-
------------------------------
...
LANL: Ice Exists on Moon. Findings Could Alter Exploration of Space http://www.abqjournal.com/scitech/1sci3-6.htm
------------------------------ Date: 04 Mar 1999 22:38:41 -0500 From: Tom Breton <tob at world.std.com> Subject: Re: Alternatives to judging Re: fx-discuss: IBM/Linux/Webserver Neal M Gafter <Neal.Gafter at Eng.Sun.Com> writes: > Tom Breton wrote: > > > On the other hand, if the coupon holders don't get to decide which > > > interpretation is used, your proposal is the same as having the judge > > > simply clarify the claim. > > > > Maybe. > > > > But you have assumed that there is still a single judge, which is not > > neccessarily true. > > I have not assumed any such thing. If you replace "judge" with > "judges" in my statement, it is still true. You said it was the same as now, and now doesn't permit multiple judges per claim. > > For another thing, it allows branching, so if a judge chooses what you > > consider the less interesting interpretation, the more interesting one > > still exists and can proceed. > > The current system does in fact allow new claims to be created; what > you are suggesting is nothing new. OK, you've got me there. > > This encourages claims to evolve in ways that stay close to the real > > question, > > What "real question" is there other than the one asked in the claim > itself? Neal, we've all seen plenty of claims that were ultimately judged on some criterion that was clearly tangential to the motivating question. > > where the current system encourages claim-authors to try to > > inscribe the claim criteria in stone, lest the claim be capriciously > > "interpreted". > > You would prefer claims be vague? I for one would not want to invest > in a claim whose meaning I could not glean. Not "be" vague, grow from vagueness to clarity without trying to reach maturity in one giant leap. I got this idea during the LunH2O discussion, after listening to other participants complain that the LunH2O numbers somehow missed the (numerical) region of interest, or something like that. Thinking about why, it occured to me that it was because the claim had no time to "grow up", it was forced to immediately be mature. > > For another thing, it allows split interpretations: If the claim has > > two or more reasonable interpretations, a judge needn't choose a > > single winner, but can divide it proportionately. > > Proportionately to what? Can divide each share proportionately between the interpretations. EG, a share of claim A turns into (say) 0.25 share in claim A-1 and 0.75 share in claim A-2. Yes, I know, it requires more precision, and granularity issues would have to be resolved. Fine. > Are you suggesting that judges should be able to create new claims > without paying a claim creation fee? The claim-creation fee is a technical detail I have put on the back burner. I see it as controlling the proliferation of silly claims. (I remember some awfully trivial claims in IF). Clearly that is not a detail I should be focussed on at this stage in this proposal. - -- Tom Breton, http://world.std.com/~tob Ugh-free Spelling (no "gh") http://world.std.com/~tob/ugh-free.html ------------------------------ ------------------------------ Date: Fri, 5 Mar 1999 10:45:27 -0700 (MST) From: "Patrik D'haeseleer" <patrik at cs.unm.edu> Subject: Re: Alternatives to judging Re: fx-discuss: IBM/Linux/Webserver Tom Breton writes: >Neal M Gafter <Neal.Gafter at Eng.Sun.Com> writes: > >> Tom Breton wrote: >> > > On the other hand, if the coupon holders don't get to decide which >> > > interpretation is used, your proposal is the same as having the judge >> > > simply clarify the claim. >> > >> > Maybe. >> > >> > But you have assumed that there is still a single judge, which is not >> > neccessarily true. >> >> I have not assumed any such thing. If you replace "judge" with >> "judges" in my statement, it is still true. > >You said it was the same as now, and now doesn't permit multiple >judges per claim. That's what the FX Outside Appeals Panel is doing. Unfortunately, it's a lot harder to get 3-4 judges to agree on how to clarify - let alone *judge* - a claim than for a single judge to do so. I think the FXOAP so far has been less than an absolute success. Each of us has other commitments in our lives than FX, and we try to get a consensus on each decision we make as a group, which requires a lot of coordination and email. If there's a single judge, he or she may be unavailable for perhaps a week, but at least can give you a concrete decision as soon as s/he gets back online. >> > This encourages claims to evolve in ways that stay close to the real >> > question, >> >> What "real question" is there other than the one asked in the claim >> itself? > >Neal, we've all seen plenty of claims that were ultimately judged on >some criterion that was clearly tangential to the motivating question. Mainly due to some individuals wanting to cling to the exact wording of the claim ;-) >> > where the current system encourages claim-authors to try to >> > inscribe the claim criteria in stone, lest the claim be capriciously >> > "interpreted". >> >> You would prefer claims be vague? I for one would not want to invest >> in a claim whose meaning I could not glean. > >Not "be" vague, grow from vagueness to clarity without trying to reach >maturity in one giant leap. > >I got this idea during the LunH2O discussion, after listening to other >participants complain that the LunH2O numbers somehow missed the >(numerical) region of interest, or something like that. Thinking >about why, it occured to me that it was because the claim had no time >to "grow up", it was forced to immediately be mature. Yes, at the time I signed up as judge, I wasn't completely satisfied with the claim wording and the leeway in judging the claim creator was willing to agree with, but I felt compelled to open the claim because there was a real chance it would become moot very soon, either by malfunction of the spacecraft, or by a dilligent, PR-oriented crew releasing the results as soon as they came in. Of course, it turned out that the spacecraft worked fine but slowly, and the Lunar Explorer team sucks in terms of releasing any data... In the future, I've decided I'd rather *not* have a claim than sign off on a too vague *or* too restrictive claim. Patrik ------------------------------
ice on Earth's Moon ? http://www.sciencemag.org/science/scripts/display/full/274/5292/1495
H2O on the Moon http://www.sciencemag.org/science/scripts/display/full/274/5292/1495
Additional informaiton about the Lunar Prospector mission can be found on the Internet at URL: http://lunar.arc.nasa.gov
Experimental Rocket Propulsion Society http://www.erps.org/ ???
Probe detects frozen water on moon http://www.msnbc.com/news/145127.asp at *both* lunar poles.
http://www.msnbc.com/news/192179.asp
"If this is really true, then the ice should be uniform layers of pure ice and not mixed in with the soil," Alan Binder, principal investigator for Lunar Prospector, told MSNBC. "If that's the case, then we're talking about maybe a billion tons of water, maybe 10 billion tons of water."
William Feldman, a Lunar Prospector co-investigator and a spectrometer specialist at the Los Alamos National Laboratory in New Mexico, estimated the amount at as much as 3 billion tons at each pole, with 15 percent more ice in the north than in the south. (Clementine detected more ice in the south, but Binder said that may have been because of the moon's slight seasonal variations.) ...
Both scientists say the ice is probably buried beneath 1 or 2 feet of lunar soil."
``Moon Water - Our Ticket to the Solar System!'' article by Dr Jamie Love 12 March 1998 http://www.synapses.co.uk/science/moonwat.html lots of details, in an easy-to-read style. Includes graphs of the actual data.
Clementine Lunar Image Browser http://www.nrl.navy.mil/clementine/clib/
Water on Mars ? http://www.spacedaily.com/news/mars-odyssey-02e.html
http://www.chron.com/content/interactive/space/astronomy/news/1999/solarsys/990722.html
July 22, 1999, 5:00 p.m. Spacecraft's collision with moon may find concrete, not water By PAUL RECER, AP Science Writer
WASHINGTON (AP) ‹ NASA is planning to smash the moon with a used spacecraft next week in a search for ice, but two scientists predict it'll turn up only a concrete-like layer of lunar soil and no water.
The $63 million Lunar Prospector mission concludes Aug. 1 and researchers plan a spectacular end for the little spacecraft -- smacking it into a constantly shaded crater on the moon's South Pole.
NASA researchers believe there is a good chance the 3,800 mph crash of the 354-pound craft will gouge a hole in the crater deep enough to free a cloud of water vapor if ice is locked in the frigid soil.
...
Lunar Prospector will smash into one such shaded crater at about 5:52 a.m. EDT on July 31.
"We selected that day because the funding for Lunar Prospector runs out on Aug. 1," said Goldstein ...
The impact could produce a vapor with as much as 40 pounds of water -- about five gallons. A cloud of vapor should become visible to powerful telescopes within four seconds after the collision. Sunlight is expected to break the water vapor into hydrogen and hydroxyl. Telescopes on Earth and in space that can detect ultraviolet radiation will be able to tell if the plume contains hydroxyl, a key marker for water.
Send comments, suggestions, bug reports to