> Jerry wrote: > > Suppose the target statistic is "the mean weight of men attending > > the upcoming commencement ceremonies at my school." I gather a > > group of ten men and find their weights vary from 131 lbs to 245 > > lbs with a mean of 162 lbs. My weights are accurate to within > > +/- 0.5 lbs, so these are genuine fluctuations in weight that I > > measure. Nevertheless, in terms of the target statistic, these > > fluctuations are noise and contribute to the error bars of the > > target statistic.
> > Your fallacy, Surfer, is assuming that since I am measuring > > genuine fluctuations in weight rather than experiencing random > > measurement error, these fluctuations in weight do not contribute > > to the error bars of the target statistic. In other words, you > > believe that the mean weight of this sample of men should be > > reported as 162 +/- 0.5 lbs.
> Yes. But remember that what is desired is NOT the mean of that > particular sample of ten men, but rather the mean of the ENTIRE set of > men attending the ceremony. It should be obvious that measuring just ten > men does not necessarily yield the same answer that measuring every man > would give.
Yes. I corrected myself immediately afterwards in a reply-to-self.
harry wrote: > Tom Roberts is a known > "pro-relativity" enthusiastic.
For the record, I am not "pro relativity", I am pro SCIENCE. But to date nobody has shown any sufficient reason to replace relativity with any other theory. And while there are indications that GR may well ultimately be replaced, within its domain there are no such indications for SR.
> I see what you mean: he gives evidence for what he calls plagiarism.
...from Christopher Bjerkness? You call this "evidence"
> > It has inasmuch as it shows Allais' bias.
> As I thought, nothing to do with his work
His "work" in the field of physics is the "work" of a crank, very close to the "qaulity" of Cahill's "work". At least Cahill is not an antisemite like Allais :-)
<Cephalobus_alie...@comcast.net> wrote: >On May 23, 6:04 am, Jerry <Cephalobus_alie...@comcast.net> wrote: >> On May 22, 10:46 pm, Surfer <n...@spam.net> wrote:
>> > Tom Robert's analysis shows that Miller's data is not consistent with >> > a model in which ether drift has a constant velocity.
>> > If one assumes such a model (as done in Robert's paper), then the >> > fluctuations in Miller's data must be attributed to measurement error, >> > and that results in huge error bars--showing that the model is not >> > supported.
>> > >and that Cahill's >> > >later work based in part on Miller's data is total fantasy.
>> > Robert's analysis fails to show this because Cahill's theory predicts >> > that the velocity of 3-space will continually vary due to >> > gravitational wave effects.
>> > If one assumes such a model, the fluctuations in Miller's data may be >> > attributed to velocity fluctuations, rather than to measurement error. >> > Hence errorbars can be as small as Miller's estimate of probable >> > errror.
>> This is the essence of your argument, as I understand it:
>> 1) Tom Roberts analyzes the data in terms of a model in which the >> measurement readings are the sum of a presumptive periodic signal >> plus random measurement error plus systematic drift. You agree >> that the data do not allow extraction of any sort of periodic >> signal from the measurement readings.
>> 2) You, on the contrary, assert that the data should be analyzed >> in terms of a model in which the measurement readings are the >> sum of a presumptive periodic signal plus random gravitational >> wave signals plus systematic drift. Under these circumstances, >> you assert that the data DO allow extraction of a periodic signal >> from the measurement readings.
>> In terms of the target statistics (the phase and amplitude of a >> presumptive periodic signal buried in the measurement readings), >> your distinction between random measurement error versus random >> gravitational wave signals is a meaningless one. You have no >> means of modeling the random gravitational wave signals, so there >> is no way of "subtracting out" their contribution to the data. >> Whether random measurement error or random signal, they >> contribute to the error bars OF THE TARGET STATISTICS.
>> Your problem is that you are confusing your target statistics. >> If your target statistics are, say, the amplitude and frequency >> distribution of the presumed gravitational wave fluctuations, >> then these statistics may be determined with high precision.
>> But these are not your target statistics. What is signal in one >> context is noise in another context.
>> Let me illustrate:
>> Suppose the target statistic is "the mean weight of men attending >> the upcoming commencement ceremonies at my school." I gather a >> group of ten men and find their weights vary from 131 lbs to 245 >> lbs with a mean of 162 lbs. My weights are accurate to within >> +/- 0.5 lbs, so these are genuine fluctuations in weight that I >> measure. Nevertheless, in terms of the target statistic, these >> fluctuations are noise and contribute to the error bars of the >> target statistic.
>> Your fallacy, Surfer, is assuming that since I am measuring >> genuine fluctuations in weight rather than experiencing random >> measurement error, these fluctuations in weight do not contribute >> to the error bars of the target statistic. In other words, you >> believe that the mean weight of this sample of men should be >> reported as 162 +/- 0.5 lbs.
>Sorry, I hit the "send" button too soon. Normally I double check >what I write before sending it.
>The mean weight of the SAMPLE of men may indeed be reported as >162 lbs with a precision limited only by the accuracy of the >scales. But the mean weight of the SAMPLE of men is used as an >estimator of the mean weight of the POPULATION. So let me rewrite >my last paragraph:
>Your fallacy, Surfer, is assuming that since I am measuring >genuine fluctuations in weight rather than experiencing random >measurement error, these fluctuations in weight do not contribute >to the error bars of the target statistic. In other words, you >believe that the mean weight of the target population of men as >estimated by my sample should be reported as 162 +/- 0.5 lbs.
My position is that:
1) Tom Roberts' analysis is invalid
2) If the velocity vector is varying due to wave effects, then it cannot be effectively sampled by a single rotation of the interferometer, because the vector will change during the course of the rotation.
3) Therefore data collected during a single rotation, should not be regarded as a complete measurement.
4) If data collected during a single rotation is not a complete measurement, then it is meaningless to apply error analysis to such data.
5) However, if a run of twenty rotations is performed, then the average values obtained at each marker MIGHT be representive of the average component of velocity in the plane of the interferomenter during the time it took to perform the rotations.
There is no way to prove whether that would be the case. The only way to find out is to try the procedure and see if sensible results are obtained.
7) Miller performed runs at different sideral times so that the plane of the interferometer would sample different components of the velocity vector as the earth turned on its axis. However, if the velocity vector is varying due to wave effects, then it cannot be effectively sampled by a single rotation of the earth, because again the vector will change during the course of the rotation.
8) Therefore velocity vector data obtained during a single rotation of the earth, should again not be regarded as a complete measurement.
9) If velocity vector data collected during a single rotation off the earth is not a complete measurement, then it would again seem meaningless to apply error analysis to such data also.
10) So I come to the conclusion that error analysis should only be applied to FINAL values for the 3-space velocity vector. I believe this was the approach of Allais.
> <Cephalobus_alie...@comcast.net> wrote: > >On May 23, 6:04 am, Jerry <Cephalobus_alie...@comcast.net> wrote:
> >> This is the essence of your argument, as I understand it:
> >> 1) Tom Roberts analyzes the data in terms of a model in which the > >> measurement readings are the sum of a presumptive periodic signal > >> plus random measurement error plus systematic drift. You agree > >> that the data do not allow extraction of any sort of periodic > >> signal from the measurement readings.
> >> 2) You, on the contrary, assert that the data should be analyzed > >> in terms of a model in which the measurement readings are the > >> sum of a presumptive periodic signal plus random gravitational > >> wave signals plus systematic drift. Under these circumstances, > >> you assert that the data DO allow extraction of a periodic signal > >> from the measurement readings.
> >> In terms of the target statistics (the phase and amplitude of a > >> presumptive periodic signal buried in the measurement readings), > >> your distinction between random measurement error versus random > >> gravitational wave signals is a meaningless one. You have no > >> means of modeling the random gravitational wave signals, so there > >> is no way of "subtracting out" their contribution to the data. > >> Whether random measurement error or random signal, they > >> contribute to the error bars OF THE TARGET STATISTICS.
> >> Your problem is that you are confusing your target statistics. > >> If your target statistics are, say, the amplitude and frequency > >> distribution of the presumed gravitational wave fluctuations, > >> then these statistics may be determined with high precision.
> >> But these are not your target statistics. What is signal in one > >> context is noise in another context.
> >> Let me illustrate:
> >> Suppose the target statistic is "the mean weight of men attending > >> the upcoming commencement ceremonies at my school." I gather a > >> group of ten men and find their weights vary from 131 lbs to 245 > >> lbs with a mean of 162 lbs. My weights are accurate to within > >> +/- 0.5 lbs, so these are genuine fluctuations in weight that I > >> measure. Nevertheless, in terms of the target statistic, these > >> fluctuations are noise and contribute to the error bars of the > >> target statistic.
> >> <snipped my goof>
> >Sorry, I hit the "send" button too soon. Normally I double check > >what I write before sending it.
> >The mean weight of the SAMPLE of men may indeed be reported as > >162 lbs with a precision limited only by the accuracy of the > >scales. But the mean weight of the SAMPLE of men is used as an > >estimator of the mean weight of the POPULATION. So let me rewrite > >my last paragraph:
> >Your fallacy, Surfer, is assuming that since I am measuring > >genuine fluctuations in weight rather than experiencing random > >measurement error, these fluctuations in weight do not contribute > >to the error bars of the target statistic. In other words, you > >believe that the mean weight of the target population of men as > >estimated by my sample should be reported as 162 +/- 0.5 lbs.
> My position is that:
> 1) Tom Roberts' analysis is invalid
You only WANT it to be invalid. None of your arguments against it hold water.
> 2) If the velocity vector is varying due to wave effects, then it > cannot be effectively sampled by a single rotation of the > interferometer, because the vector will change during the course of > the rotation.
> 3) Therefore data collected during a single rotation, should not be > regarded as a complete measurement.
> 4) If data collected during a single rotation is not a complete > measurement, then it is meaningless to apply error analysis to such > data.
Who do you accuse of attempting that? The above critique applies to no one in this group.
> 5) However, if a run of twenty rotations is performed, then the > average values obtained at each marker MIGHT be representive of the > average component of velocity in the plane of the interferomenter > during the time it took to perform the rotations.
> There is no way to prove whether that would be the case. The only way > to find out is to try the procedure and see if sensible results are > obtained.
Unfortunately, your only criterion for whether the results are "sensible" seems to be whether the results fit your preconceived prejudices. You reject objective statistical analysis.
> 7) Miller performed runs at different sideral times so that the plane > of the interferometer would sample different components of the > velocity vector as the earth turned on its axis. However, if the > velocity vector is varying due to wave effects, then it cannot be > effectively sampled by a single rotation of the earth, because again > the vector will change during the course of the rotation.
> 8) Therefore velocity vector data obtained during a single rotation of > the earth, should again not be regarded as a complete measurement.
> 9) If velocity vector data collected during a single rotation off the > earth is not a complete measurement, then it would again seem > meaningless to apply error analysis to such data also.
> 10) So I come to the conclusion that error analysis should only be > applied to FINAL values for the 3-space velocity vector. I believe > this was the approach of Allais.
WHAAAT???
Miller's final values represent an executive summary of thousands of measurements, processed using data reduction methods that are invalid by modern standards.
His processed data are like company annual reports before the passage of the Sarbanes-Oxley act. Being from Australia, you probably wouldn't know of Sarbanes-Oxley, but I'm sure you HAVE heard of the financial scandals that led to its passage: Enron, Tyco International, WorldCom, Adelphia, Peregrine Systems etc.
Even an astute investor examining the Enron annual report would never have guessed at the company's weakness. However, had the said astute investor been privy to inside financial information, such as the sort of information that Arthur Andersen had access to but did not examine properly, said astute investor wouldn't have touched Enron with a ten-foot pole.
Tom Roberts has access to Miller's raw data. His analysis of the data available to him shows that Miller's final results are not justified. By analogy, Tom Roberts is in the position of the astute investor with inside access to Enron's books. He isn't buying, and I'm not buying either.
> harry wrote: >> Tom Roberts is a known "pro-relativity" enthusiastic.
> For the record, I am not "pro relativity", I am pro SCIENCE. But to date > nobody has shown any sufficient reason to replace relativity with any > other theory. And while there are indications that GR may well ultimately > be replaced, within its domain there are no such indications for SR.
Precision appreciated. :-) Of course I was simplifying (Allais is similarly pro SCIENCE) in a vain attempt to steer "Dono" away from talking as if science is like a fan club. But it didn't help, and I should have known...
> Jerry wrote: [...] > Imagine that one selected a different group of ten men. The average for > this second group is almost surely not 162 lbs. Consider a third, fourth, > fifth,... group of ten men, and plot the distriution of the averages for > the different groups. The AVERAGES will display a variance, and that > variance is related to the variance of the weights of the individual men. > THIS is what statistics does: it tells you what the variance of the > average will be, given the variance of the individual measurements (here > mens' weights).
> For the case (like Miller's) where you have only one group of ten men to > consider, honesty precludes one from claiming the average is 162 +- 0.5, > and one must claim 162 +- sigma, where sigma is determined from the > distribution of the ten mens' weights. In the language of statistics, the > mean of those ten mens' weights is the best unbiased predictor of the true > average, and the sigma is the best unbiased predictor of how accurately > the average of those ten weights reflects the true average. Note these are > "predictors", because one does not know the true values, one only knows > the ten values one measured.
> To learn how to compute that sigma you need to STUDY. If > those ten men's weights are randomly but uniformly > distributed between 131 and 245 lbs, the sigma (errorbar > on the average) will be about 10 lbs, not 0.5 lbs.
> That's PRECISELY what I did for each run of Miller's data: For each of his > eight orientations he averaged 40 data points. I computed the variance of > those eight averages from the variance of the 40 points that went into > computing each one. Those variances (errorbars) GREATLY exceed the > variation among the eight averages, showing that the variation Miller used > to make his result is not significant. This, in turn, makes any conclusion > based on his results be insignificant: Miller concluded the average is 11 > km/s, but the errorbar on that average is something like 100 km/s; Miller > determined an average direction, but the errorbar on that direction > includes all possible directions.
| [...] | > Imagine that one selected a different group of ten men. The average for | > this second group is almost surely not 162 lbs. Consider a third, fourth, | > fifth,... group of ten men, and plot the distriution of the averages for | > the different groups. The AVERAGES will display a variance, and that | > variance is related to the variance of the weights of the individual men. | > THIS is what statistics does: it tells you what the variance of the | > average will be, given the variance of the individual measurements (here | > mens' weights). | > | > For the case (like Miller's) where you have only one group of ten men to | > consider, honesty precludes one from claiming the average is 162 +- 0.5, | > and one must claim 162 +- sigma, where sigma is determined from the | > distribution of the ten mens' weights. In the language of statistics, the | > mean of those ten mens' weights is the best unbiased predictor of the true | > average, and the sigma is the best unbiased predictor of how accurately | > the average of those ten weights reflects the true average. Note these are | > "predictors", because one does not know the true values, one only knows | > the ten values one measured. | > | > To learn how to compute that sigma you need to STUDY. If | > those ten men's weights are randomly but uniformly | > distributed between 131 and 245 lbs, the sigma (errorbar | > on the average) will be about 10 lbs, not 0.5 lbs. | > | > That's PRECISELY what I did for each run of Miller's data: For each of his | > eight orientations he averaged 40 data points. I computed the variance of | > those eight averages from the variance of the 40 points that went into | > computing each one. Those variances (errorbars) GREATLY exceed the | > variation among the eight averages, showing that the variation Miller used | > to make his result is not significant. This, in turn, makes any conclusion | > based on his results be insignificant: Miller concluded the average is 11 | > km/s, but the errorbar on that average is something like 100 km/s; Miller | > determined an average direction, but the errorbar on that direction | > includes all possible directions. | | Well explained this time! :-) | | Harald
Consider why did Einstein say the speed of light from A to B is c-v, the speed of light from B to A is c+v, the "time" each way is the same, and tell us what the fucking errorbar is, you handwaving ignorant arse-kissing prat.
> "harry" <harald.vanlintelButNotT...@epfl.ch> wrote in message > news:4837e477$1_5@news.bluewin.ch... > | > | "Tom Roberts" <tjroberts...@sbcglobal.net> wrote in message > | news:99BZj.108$uE5.87@flpi144.ffdc.sbc.com... > | > Jerry wrote: > | [...] > | > Imagine that one selected a different group of ten men. The average > for > | > this second group is almost surely not 162 lbs. Consider a third, > fourth, > | > fifth,... group of ten men, and plot the distriution of the averages > for > | > the different groups. The AVERAGES will display a variance, and that > | > variance is related to the variance of the weights of the individual > men. > | > THIS is what statistics does: it tells you what the variance of the > | > average will be, given the variance of the individual measurements > (here > | > mens' weights). > | > > | > For the case (like Miller's) where you have only one group of ten men > to > | > consider, honesty precludes one from claiming the average is 162 +- > 0.5, > | > and one must claim 162 +- sigma, where sigma is determined from the > | > distribution of the ten mens' weights. In the language of statistics, > the > | > mean of those ten mens' weights is the best unbiased predictor of the > true > | > average, and the sigma is the best unbiased predictor of how > accurately > | > the average of those ten weights reflects the true average. Note these > are > | > "predictors", because one does not know the true values, one only > knows > | > the ten values one measured. > | > > | > To learn how to compute that sigma you need to STUDY. If > | > those ten men's weights are randomly but uniformly > | > distributed between 131 and 245 lbs, the sigma (errorbar > | > on the average) will be about 10 lbs, not 0.5 lbs. > | > > | > That's PRECISELY what I did for each run of Miller's data: For each of > his > | > eight orientations he averaged 40 data points. I computed the variance > of > | > those eight averages from the variance of the 40 points that went into > | > computing each one. Those variances (errorbars) GREATLY exceed the > | > variation among the eight averages, showing that the variation Miller > used > | > to make his result is not significant. This, in turn, makes any > conclusion > | > based on his results be insignificant: Miller concluded the average is > 11 > | > km/s, but the errorbar on that average is something like 100 km/s; > Miller > | > determined an average direction, but the errorbar on that direction > | > includes all possible directions. > | > | Well explained this time! :-) > | > | Harald > Consider why did Einstein say > the speed of light from A to B is c-v, > the speed of light from B to A is c+v, > the "time" each way is the same,
Easy: he did NOT say that.
> and tell us what the fucking errorbar is,
Also easy: in such theoretical approaches there isn't any. And no need to curse - that won't help you to understand it!
> you handwaving ignorant arse-kissing prat.
If you had the slightest reading ability, you would not say such silly things. :-)
How so, Surfer? Let me repeat some previous words of mine, followed by Tom's extension of my remarks. I challenge you to come up with a valid critique of Tom's procedure for determining errorbars.
----------------------------------------------------------------- (Director's Cut, replacing my blooper with my corrected remarks)
I wrote:
Suppose the target statistic is "the mean weight of men attending the upcoming commencement ceremonies at my school." I gather a group of ten men and find their weights vary from 131 lbs to 245 lbs with a mean of 162 lbs. My weights are accurate to within +/- 0.5 lbs, so these are genuine fluctuations in weight that I measure. Nevertheless, in terms of the target statistic, these fluctuations are noise and contribute to the error bars of the target statistic.
Your fallacy, Surfer, is assuming that since I am measuring genuine fluctuations in weight rather than experiencing random measurement error, these fluctuations in weight do not contribute to the error bars of the target statistic. In other words, you believe that the mean weight of the target population of men as estimated by my sample should be reported as 162 +/- 0.5 lbs.
Imagine that one selected a different group of ten men. The average for this second group is almost surely not 162 lbs. Consider a third, fourth, fifth,... group of ten men, and plot the distriution of the averages for the different groups. The AVERAGES will display a variance, and that variance is related to the variance of the weights of the individual men. THIS is what statistics does: it tells you what the variance of the average will be, given the variance of the individual measurements (here mens' weights).
For the case (like Miller's) where you have only one group of ten men to consider, honesty precludes one from claiming the average is 162 +- 0.5, and one must claim 162 +- sigma, where sigma is determined from the distribution of the ten mens' weights. In the language of statistics, the mean of those ten mens' weights is the best unbiased predictor of the true average, and the sigma is the best unbiased predictor of how accurately the average of those ten weights reflects the true average. Note these are "predictors", because one does not know the true values, one only knows the ten values one measured.
To learn how to compute that sigma you need to STUDY. If those ten men's weights are randomly but uniformly distributed between 131 and 245 lbs, the sigma (errorbar on the average) will be about 10 lbs, not 0.5 lbs.
That's PRECISELY what I did for each run of Miller's data: For each of his eight orientations he averaged 40 data points. I computed the variance of those eight averages from the variance of the 40 points that went into computing each one. Those variances (errorbars) GREATLY exceed the variation among the eight averages, showing that the variation Miller used to make his result is not significant. This, in turn, makes any conclusion based on his results be insignificant: Miller concluded the average is 11 km/s, but the errorbar on that average is something like 100 km/s; Miller determined an average direction, but the errorbar on that direction includes all possible directions.
Surfer wrote: > My position is that: > 1) Tom Roberts' analysis is invalid
That's an ABSURD "position", as it implies you don't believe in elementary mathematics.
[I am only discussing the error analysis of Miller's data.]
> 2) If the velocity vector is varying due to wave effects, then it > cannot be effectively sampled by a single rotation of the > interferometer, because the vector will change during the course of > the rotation.
Then Miller's entire approach is invalidated. So you cannot use his result.
> 3) Therefore data collected during a single rotation, should not be > regarded as a complete measurement.
This is in conflict with your (2) -- if the velocity vector is varying significantly within the 20 seconds of a turn, then the entire measurement approach is invalidated, it's not merely an "incomplete measurement".
> 4) If data collected during a single rotation is not a complete > measurement, then it is meaningless to apply error analysis to such > data.
But it is NOT meaningless to apply error analysis to MILLER'S ANALYSIS or his conclusion.
> 5) However, if a run of twenty rotations is performed, then the > average values obtained at each marker MIGHT be representive of the > average component of velocity in the plane of the interferomenter > during the time it took to perform the rotations.
OK. But such an average is only "representative" to the accuracy of an error analysis performed n the data. That is, for Miller it is NOT significant, and the AVERAGE is fully consistent with zero.
> There is no way to prove whether that would be the case. The only way > to find out is to try the procedure and see if sensible results are > obtained.
Sure there is! -- Perform the error analysis and look for statistical significance. that fact that his average values are not significantly different from zero means your hopes and dreams are not realized by Miller's measurements.
> 7) Miller performed runs at different sideral times so that the plane > of the interferometer would sample different components of the > velocity vector as the earth turned on its axis. However, if the > velocity vector is varying due to wave effects, then it cannot be > effectively sampled by a single rotation of the earth, because again > the vector will change during the course of the rotation.
See above -- an error analysis will still tell you whether or not the result is significant.
> 10) So I come to the conclusion that error analysis should only be > applied to FINAL values for the 3-space velocity vector. I believe > this was the approach of Allais.
Allais screwed up but did not realize it. So did Cahill. So do you. <shrug>
> : His "work" in the field of physics is the "work" of a crank, very > : close to the "qaulity" of Cahill's "work".
> Nasa disagrees; please show his calculation errors.
Both Allais and Cahill arrived to the same conclusion after:
-not running any experiment -reanalizing the Miller data
...that they could detect absolute motion. This is pure crank material.
> : At least Cahill is not an antisemite like Allais :-)
> ??! That is a strong accusation, which makes you "anti-Allais" according to > your use of the term - and it's just as irrelevant.
> Harald
Don't try to deflect the discussion, the point is about what Allais writes on his website, it is pretty obvious that he's an antisemite. This colors his "scientific" work.
On May 24, 1:21 am, "harry" <harald.vanlintelButNotT...@epfl.ch> wrote:
> : His "work" in the field of physics is the "work" of a crank, very > : close to the "qaulity" of Cahill's "work".
> Nasa disagrees; please show his calculation errors.
Really? Nasa agrees with Allais' crackpottery on the Miller experiment? You have always been very good with references , Harald, can you give at least one reference that shows that Nasa agrees with Allais on the detection of absolute motion?
| "Androcles" <Headmas...@Hogwarts.physics> wrote in message
| news:RQRZj.103193$AN7.87552@newsfe23.ams2... | > | > "harry" <harald.vanlintelButNotT...@epfl.ch> wrote in message | > news:4837e477$1_5@news.bluewin.ch... | > | | > | "Tom Roberts" <tjroberts...@sbcglobal.net> wrote in message | > | news:99BZj.108$uE5.87@flpi144.ffdc.sbc.com... | > | > Jerry wrote: | > | [...] | > | > Imagine that one selected a different group of ten men. The average | > for | > | > this second group is almost surely not 162 lbs. Consider a third, | > fourth, | > | > fifth,... group of ten men, and plot the distriution of the averages | > for | > | > the different groups. The AVERAGES will display a variance, and that | > | > variance is related to the variance of the weights of the individual | > men. | > | > THIS is what statistics does: it tells you what the variance of the | > | > average will be, given the variance of the individual measurements | > (here | > | > mens' weights). | > | > | > | > For the case (like Miller's) where you have only one group of ten men | > to | > | > consider, honesty precludes one from claiming the average is 162 +- | > 0.5, | > | > and one must claim 162 +- sigma, where sigma is determined from the | > | > distribution of the ten mens' weights. In the language of statistics, | > the | > | > mean of those ten mens' weights is the best unbiased predictor of the | > true | > | > average, and the sigma is the best unbiased predictor of how | > accurately | > | > the average of those ten weights reflects the true average. Note these | > are | > | > "predictors", because one does not know the true values, one only | > knows | > | > the ten values one measured. | > | > | > | > To learn how to compute that sigma you need to STUDY. If | > | > those ten men's weights are randomly but uniformly | > | > distributed between 131 and 245 lbs, the sigma (errorbar | > | > on the average) will be about 10 lbs, not 0.5 lbs. | > | > | > | > That's PRECISELY what I did for each run of Miller's data: For each of | > his | > | > eight orientations he averaged 40 data points. I computed the variance | > of | > | > those eight averages from the variance of the 40 points that went into | > | > computing each one. Those variances (errorbars) GREATLY exceed the | > | > variation among the eight averages, showing that the variation Miller | > used | > | > to make his result is not significant. This, in turn, makes any | > conclusion | > | > based on his results be insignificant: Miller concluded the average is | > 11 | > | > km/s, but the errorbar on that average is something like 100 km/s; | > Miller | > | > determined an average direction, but the errorbar on that direction | > | > includes all possible directions. | > | | > | Well explained this time! :-) | > | | > | Harald | | > Consider why did Einstein say | > the speed of light from A to B is c-v, | > the speed of light from B to A is c+v, | > the "time" each way is the same, | | Easy: he did NOT say that.
Yes he did, you stupid lying prat. Quote: "we establish by definition that the ``time'' required by light to travel from A to B equals the ``time'' it requires to travel from B to A. " Unquote. Quote: "But the ray moves relatively to the initial point of k, when measured in the stationary system, with the velocity c-v". Unquote.
Where do you imagine the cuckoo malformations come from, thin aether?
| > and tell us what the fucking errorbar is, | | Also easy: in such theoretical approaches there isn't any. And no need to | curse - that won't help you to understand it!
| | > you handwaving ignorant arse-kissing prat. | | If you had the slightest reading ability, you would not say such silly | things. :-) | If YOU has the slightest reading ability you would not LIE about what Einstein said, you ignorant fucking tord who loves to kiss the clown Roberts' arse.
-- Why did Einstein say the speed of light from A to B is c-v, the speed of light from B to A is c+v, the "time" each way is the same?
ahahahaha... AHAHAHAHA... AHAHAHAHAHA.... Andro, you are true preacher of hail & brimstone relativity... ahahahaha.. But consider that Einstein Dingleberries are damaged goods who are incapable of recognizing that == Relativism, like Religion, is just a tool that is used === by the few to fuck the many. ==== Nobody is born religious nor relativious. ===== SR/GR, like Religion, is an acquired disease. ====== AND/OR SR/GR, like Religion, is Opium for the ========= Einstein Dingleberries. --- Halleluiah! ---
Thanks for the laughs from your exchange with the Swiss cheese guy below. ahahaha... ahahahanson
"Androcles" <Headmas...@Hogwarts.physics> wrote in message
> "harry" <harald.vanlintelButNotT...@epfl.ch> wrote in message > news:4837fa4b$1_6@news.bluewin.ch... > | > | "Androcles" <Headmas...@Hogwarts.physics> wrote in message > | news:RQRZj.103193$AN7.87552@newsfe23.ams2... > | > > | > "harry" <harald.vanlintelButNotT...@epfl.ch> wrote in message > | > news:4837e477$1_5@news.bluewin.ch... > | > | > | > | "Tom Roberts" <tjroberts...@sbcglobal.net> wrote in message > | > | news:99BZj.108$uE5.87@flpi144.ffdc.sbc.com... > | > | > Jerry wrote: > | > | [...] > | > | > Imagine that one selected a different group of ten men. The > average > | > for > | > | > this second group is almost surely not 162 lbs. Consider a third, > | > fourth, > | > | > fifth,... group of ten men, and plot the distriution of the > averages > | > for > | > | > the different groups. The AVERAGES will display a variance, and > that > | > | > variance is related to the variance of the weights of the > individual > | > men. > | > | > THIS is what statistics does: it tells you what the variance of > the > | > | > average will be, given the variance of the individual measurements > | > (here > | > | > mens' weights). > | > | > > | > | > For the case (like Miller's) where you have only one group of ten > men > | > to > | > | > consider, honesty precludes one from claiming the average is 162 > +- > | > 0.5, > | > | > and one must claim 162 +- sigma, where sigma is determined from > the > | > | > distribution of the ten mens' weights. In the language of > statistics, > | > the > | > | > mean of those ten mens' weights is the best unbiased predictor of > the > | > true > | > | > average, and the sigma is the best unbiased predictor of how > | > accurately > | > | > the average of those ten weights reflects the true average. Note > these > | > are > | > | > "predictors", because one does not know the true values, one only > | > knows > | > | > the ten values one measured. > | > | > > | > | > To learn how to compute that sigma you need to STUDY. If > | > | > those ten men's weights are randomly but uniformly > | > | > distributed between 131 and 245 lbs, the sigma (errorbar > | > | > on the average) will be about 10 lbs, not 0.5 lbs. > | > | > > | > | > That's PRECISELY what I did for each run of Miller's data: For > each > of > | > his > | > | > eight orientations he averaged 40 data points. I computed the > variance > | > of > | > | > those eight averages from the variance of the 40 points that went > into > | > | > computing each one. Those variances (errorbars) GREATLY exceed the > | > | > variation among the eight averages, showing that the variation > Miller > | > used > | > | > to make his result is not significant. This, in turn, makes any > | > conclusion > | > | > based on his results be insignificant: Miller concluded the > average > is > | > 11 > | > | > km/s, but the errorbar on that average is something like 100 km/s; > | > Miller > | > | > determined an average direction, but the errorbar on that > direction > | > | > includes all possible directions. > | > | > | > | Well explained this time! :-) > | > | > | > | Harald > | > | > Consider why did Einstein say > | > the speed of light from A to B is c-v, > | > the speed of light from B to A is c+v, > | > the "time" each way is the same, > | > | Easy: he did NOT say that.
> Yes he did, you stupid lying prat. > Quote: > "we establish by definition that the ``time'' required by light to travel > from A to B equals the ``time'' it requires to travel from B to A. " > Unquote. > Quote: > "But the ray moves relatively to the initial point of k, when measured in > the stationary system, with the velocity c-v". > Unquote.
> Where do you imagine the cuckoo malformations come from, thin aether?
> | > and tell us what the fucking errorbar is, > | > | Also easy: in such theoretical approaches there isn't any. And no need > to > | curse - that won't help you to understand it!
> | > | > you handwaving ignorant arse-kissing prat. > | > | If you had the slightest reading ability, you would not say such silly > | things. :-) > | > If YOU has the slightest reading ability you would not LIE about what > Einstein > said, you ignorant fucking tord who loves to kiss the clown Roberts' arse.
> -- > Why did Einstein say > the speed of light from A to B is c-v, > the speed of light from B to A is c+v, > the "time" each way is the same?
<Cephalobus_alie...@comcast.net> wrote: >On May 23, 9:30 pm, Surfer <n...@spam.net> wrote: >> On Fri, 23 May 2008 04:16:38 -0700 (PDT), Jerry
>> <Cephalobus_alie...@comcast.net> wrote: >> >On May 23, 6:04 am, Jerry <Cephalobus_alie...@comcast.net> wrote:
>> >> This is the essence of your argument, as I understand it:
>> >> 1) Tom Roberts analyzes the data in terms of a model in which the >> >> measurement readings are the sum of a presumptive periodic signal >> >> plus random measurement error plus systematic drift. You agree >> >> that the data do not allow extraction of any sort of periodic >> >> signal from the measurement readings.
>> >> 2) You, on the contrary, assert that the data should be analyzed >> >> in terms of a model in which the measurement readings are the >> >> sum of a presumptive periodic signal plus random gravitational >> >> wave signals plus systematic drift. Under these circumstances, >> >> you assert that the data DO allow extraction of a periodic signal >> >> from the measurement readings.
>> >> In terms of the target statistics (the phase and amplitude of a >> >> presumptive periodic signal buried in the measurement readings), >> >> your distinction between random measurement error versus random >> >> gravitational wave signals is a meaningless one. You have no >> >> means of modeling the random gravitational wave signals, so there >> >> is no way of "subtracting out" their contribution to the data. >> >> Whether random measurement error or random signal, they >> >> contribute to the error bars OF THE TARGET STATISTICS.
>> >> Your problem is that you are confusing your target statistics. >> >> If your target statistics are, say, the amplitude and frequency >> >> distribution of the presumed gravitational wave fluctuations, >> >> then these statistics may be determined with high precision.
>> >> But these are not your target statistics. What is signal in one >> >> context is noise in another context.
>> >> Let me illustrate:
>> >> Suppose the target statistic is "the mean weight of men attending >> >> the upcoming commencement ceremonies at my school." I gather a >> >> group of ten men and find their weights vary from 131 lbs to 245 >> >> lbs with a mean of 162 lbs. My weights are accurate to within >> >> +/- 0.5 lbs, so these are genuine fluctuations in weight that I >> >> measure. Nevertheless, in terms of the target statistic, these >> >> fluctuations are noise and contribute to the error bars of the >> >> target statistic.
>> >> <snipped my goof>
>> >Sorry, I hit the "send" button too soon. Normally I double check >> >what I write before sending it.
>> >The mean weight of the SAMPLE of men may indeed be reported as >> >162 lbs with a precision limited only by the accuracy of the >> >scales. But the mean weight of the SAMPLE of men is used as an >> >estimator of the mean weight of the POPULATION. So let me rewrite >> >my last paragraph:
>> >Your fallacy, Surfer, is assuming that since I am measuring >> >genuine fluctuations in weight rather than experiencing random >> >measurement error, these fluctuations in weight do not contribute >> >to the error bars of the target statistic. In other words, you >> >believe that the mean weight of the target population of men as >> >estimated by my sample should be reported as 162 +/- 0.5 lbs.
>> My position is that:
>> 1) Tom Roberts' analysis is invalid
>You only WANT it to be invalid.
It has nothing to do with what I want. His paper contains false premises.
FALSE PREMISE 1 ===================== The caption under Fig 3. says:
"The assumed-linear systematic drift from the data of Fig. 1. The lines are between successive Marker 1 values and the points are Marker 9. These markers are 180 degrees apart, so any real signal has the same value for every corner and every point--the variations are purely an instrumentation effect."
This statement is FALSE, because measurements at Marker 1 and Marker 9 were not made simultaneously. So any real FLUCTUATING signal would have different values at the two markers.
Consequently the analysis that relies on the statement is FALSE. ====================== FALSE PREMISE 2
At the top of page 6, Tom Roberts wrote:
data = signal(orientation) + systematic(time)
The key point is that signal(orientation) is independent of time, and for each orientation (marker) it has the same value for every turn of the interferometer within a given data run Therefore if the data from the first turn is subtracted marker-by-marker from the data of every turn, the result is completely independent of any orientation dependence, and contains only systematic(time).
The above claims are false, because any real FLUCTUATING signal would vary with orientation AND TIME.
So in particular, the claim that:
"Therefore if the data from the first turn is subtracted marker-by-marker from the data of every turn, the result is completely independent of any orientation dependence, and contains only systematic(time)."
is FALSE.
Consequently all analysis that relies on the claim is FALSE. =======================
>None of your arguments against it hold water.
Since you have not shown how false premises could turned into true ones, you seem to be in denial.
>> 2) If the velocity vector is varying due to wave effects, then it >> cannot be effectively sampled by a single rotation of the >> interferometer, because the vector will change during the course of >> the rotation.
>> 3) Therefore data collected during a single rotation, should not be >> regarded as a complete measurement.
>> 4) If data collected during a single rotation is not a complete >> measurement, then it is meaningless to apply error analysis to such >> data.
you will see that all the data is from single rotations and single marker readings.
Non of this data amounts to a complete measurement of a velocity value.
>> 5) However, if a run of twenty rotations is performed, then the >> average values obtained at each marker MIGHT be representive of the >> average component of velocity in the plane of the interferomenter >> during the time it took to perform the rotations.
>> There is no way to prove whether that would be the case. The only way >> to find out is to try the procedure and see if sensible results are >> obtained.
>Unfortunately, your only criterion for whether the results are >"sensible" seems to be whether the results fit your preconceived >prejudices. You reject objective statistical analysis.
The only statistical analysis I am currently rejecting is analysis by Tom Roberts that is based on wrong premises.
>> 7) Miller performed runs at different sideral times so that the plane >> of the interferometer would sample different components of the >> velocity vector as the earth turned on its axis. However, if the >> velocity vector is varying due to wave effects, then it cannot be >> effectively sampled by a single rotation of the earth, because again >> the vector will change during the course of the rotation.
>> 8) Therefore velocity vector data obtained during a single rotation of >> the earth, should again not be regarded as a complete measurement.
>> 9) If velocity vector data collected during a single rotation off the >> earth is not a complete measurement, then it would again seem >> meaningless to apply error analysis to such data also.
>> 10) So I come to the conclusion that error analysis should only be >> applied to FINAL values for the 3-space velocity vector. I believe >> this was the approach of Allais.
>WHAAAT???
>Miller's final values represent an executive summary of thousands >of measurements, processed using data reduction methods that are >invalid by modern standards.
If that were true, Allais would have noticed.
>Tom Roberts has access to Miller's raw data. His analysis of the >data available to him shows that Miller's final results are not >justified.
> It has nothing to do with what I want. > His paper contains false premises.
You are in broken record mode, Surfer. All of your points have either been previously answered, or are irrelevant "chaff" arguments thrown up to distract attention from the weaknesses of Miller's paper. So I am snipping.
[BIG SNIP]
You have NOT, however, adequately answered Tom or myself on the following. I challenge you to come up with a valid critique of Tom's procedure for determining errorbars.
If you cannot refute this procedure of elementary statistical analysis, then Miller's entire thesis falls apart.
I will begin with a "Director's Cut" quotation from myself, (replacing my blooper with my corrected remarks) followed by a quote from Tom's post:
Suppose the target statistic is "the mean weight of men attending the upcoming commencement ceremonies at my school." I gather a group of ten men and find their weights vary from 131 lbs to 245 lbs with a mean of 162 lbs. My weights are accurate to within +/- 0.5 lbs, so these are genuine fluctuations in weight that I measure. Nevertheless, in terms of the target statistic, these fluctuations are noise and contribute to the error bars of the target statistic.
Your fallacy, Surfer, is assuming that since I am measuring genuine fluctuations in weight rather than experiencing random measurement error, these fluctuations in weight do not contribute to the error bars of the target statistic. In other words, you believe that the mean weight of the target population of men as estimated by my sample should be reported as 162 +/- 0.5 lbs.
----------------------------------------------------------------- Tom replied:
Imagine that one selected a different group of ten men. The average for this second group is almost surely not 162 lbs. Consider a third, fourth, fifth,... group of ten men, and plot the distriution of the averages for the different groups. The AVERAGES will display a variance, and that variance is related to the variance of the weights of the individual men. THIS is what statistics does: it tells you what the variance of the average will be, given the variance of the individual measurements (here mens' weights).
For the case (like Miller's) where you have only one group of ten men to consider, honesty precludes one from claiming the average is 162 +- 0.5, and one must claim 162 +- sigma, where sigma is determined from the distribution of the ten mens' weights. In the language of statistics, the mean of those ten mens' weights is the best unbiased predictor of the true average, and the sigma is the best unbiased predictor of how accurately the average of those ten weights reflects the true average. Note these are "predictors", because one does not know the true values, one only knows the ten values one measured.
To learn how to compute that sigma you need to STUDY. If those ten men's weights are randomly but uniformly distributed between 131 and 245 lbs, the sigma (errorbar on the average) will be about 10 lbs, not 0.5 lbs.
That's PRECISELY what I did for each run of Miller's data: For each of his eight orientations he averaged 40 data points. I computed the variance of those eight averages from the variance of the 40 points that went into computing each one. Those variances (errorbars) GREATLY exceed the variation among the eight averages, showing that the variation Miller used to make his result is not significant. This, in turn, makes any conclusion based on his results be insignificant: Miller concluded the average is 11 km/s, but the errorbar on that average is something like 100 km/s; Miller determined an average direction, but the errorbar on that direction includes all possible directions.
FALSE PREMISE 1 ===================== The caption under Fig 3. says:
"The assumed-linear systematic drift from the data of Fig. 1. The lines are between successive Marker 1 values and the points are Marker 9. These markers are 180 degrees apart, so any real signal has the same value for every corner and every point--the variations are purely an instrumentation effect."
This statement is FALSE, because measurements at Marker 1 and Marker 9 were not made simultaneously. So any real FLUCTUATING signal would have different values at the two markers.
Consequently the analysis that relies on the statement is FALSE. ====================== FALSE PREMISE 2
At the top of page 6, Tom Roberts wrote:
data = signal(orientation) + systematic(time)
The key point is that signal(orientation) is independent of time, and for each orientation (marker) it has the same value for every turn of the interferometer within a given data run Therefore if the data from the first turn is subtracted marker-by-marker from the data of every turn, the result is completely independent of any orientation dependence, and contains only systematic(time).
The above claims are false, because any real FLUCTUATING signal would vary with orientation AND TIME.
So in particular, the claim that:
"Therefore if the data from the first turn is subtracted marker-by-marker from the data of every turn, the result is completely independent of any orientation dependence, and contains only systematic(time)."
is FALSE.
Consequently all analysis that relies on the claim is FALSE. =======================
> [I am only discussing the error analysis of Miller's data.]
>> 2) If the velocity vector is varying due to wave effects, then it >> cannot be effectively sampled by a single rotation of the >> interferometer, because the vector will change during the course of >> the rotation.
>Then Miller's entire approach is invalidated. So you cannot use his result.
I think it is more correct to say that the original theory justifying the approach is invalidated. That doesn't rule out the possibility of finding a new theory to justify the approach.
Also I think the validity of the approach really depends on whether or not it works in practice.
Since Miller's results were found to be valid by Allais and Cahill, it seems that Miller's approach did work in practice.
>> 3) Therefore data collected during a single rotation, should not be >> regarded as a complete measurement.
>This is in conflict with your (2) -- if the velocity vector is varying >significantly within the 20 seconds of a turn, then the entire >measurement approach is invalidated, it's not merely an "incomplete >measurement".
>> 4) If data collected during a single rotation is not a complete >> measurement, then it is meaningless to apply error analysis to such >> data.
>But it is NOT meaningless to apply error analysis to MILLER'S ANALYSIS >or his conclusion.
>> 5) However, if a run of twenty rotations is performed, then the >> average values obtained at each marker MIGHT be representive of the >> average component of velocity in the plane of the interferomenter >> during the time it took to perform the rotations.
>OK. But such an average is only "representative" to the accuracy of an >error analysis performed n the data. That is, for Miller it is NOT >significant, and the AVERAGE is fully consistent with zero.
Only if you believe the above false premises.
>> There is no way to prove whether that would be the case. The only way >> to find out is to try the procedure and see if sensible results are >> obtained.
>Sure there is! -- Perform the error analysis and look for statistical >significance.
I find it hard to see how a statistical error analysis of unprocessed partial values of a quantity could tell us anything useful about the final processed value.
Suppose our equipment can only separately measure hundreds, tens and units, so prior to processing we have the partial values 5, 9, 8 and 6, 1, 2.
The final processed values differ by only 7.1%.
In contrast, the unprocessed partial values differ by 10%, -80% and -60% of their full range. This tells us nothing useful about the validity of the final processed value.
Similarly, in the case of the Miller experiments, a small change in direction of the velocity vector can produce a huge change in the components measured by the interferometer.
>different from zero means your hopes and dreams are not realized by >Miller's measurements.
>> 7) Miller performed runs at different sideral times so that the plane >> of the interferometer would sample different components of the >> velocity vector as the earth turned on its axis. However, if the >> velocity vector is varying due to wave effects, then it cannot be >> effectively sampled by a single rotation of the earth, because again >> the vector will change during the course of the rotation.
>See above -- an error analysis will still tell you whether or not the >result is significant.
>> 10) So I come to the conclusion that error analysis should only be >> applied to FINAL values for the 3-space velocity vector. I believe >> this was the approach of Allais.
>Allais screwed up but did not realize it. So did Cahill. So do you. <shrug>
If you can quote anything from their papers that contains a clear error I would like to see it.
On Sun, 25 May 2008 12:57:48 +0930, Surfer <n...@spam.net> wrote:
>I find it hard to see how a statistical error analysis of unprocessed >partial values of a quantity could tell us anything useful about the >final processed value.
>Suppose our equipment can only separately measure hundreds, tens and >units, so prior to processing we have the partial values 5, 9, 8 and >6, 1, 2.
Which after processing will be 598 and 612. Sorry if that was not clear.
>The final processed values differ by only 7.1%.
>In contrast, the unprocessed partial values differ by 10%, -80% and >-60% of their full range. This tells us nothing useful about the >validity of the final processed value.
>Similarly, in the case of the Miller experiments, a small change in >direction of the velocity vector can produce a huge change in the >components measured by the interferometer.
>So I'd say the same kind of problem applies here.
<Cephalobus_alie...@comcast.net> wrote: >On May 24, 7:34 pm, Surfer <n...@spam.net> wrote:
>> It has nothing to do with what I want. >> His paper contains false premises.
>You are in broken record mode, Surfer. All of your points have >either been previously answered, or are irrelevant "chaff" >arguments thrown up to distract attention from the weaknesses >of Miller's paper. So I am snipping.
> [BIG SNIP]
>You have NOT, however, adequately answered Tom or myself on the >following. I challenge you to come up with a valid critique of >Tom's procedure for determining errorbars.
>If you cannot refute this procedure of elementary statistical >analysis, then Miller's entire thesis falls apart.
>I will begin with a "Director's Cut" quotation from myself, >(replacing my blooper with my corrected remarks) followed by a >quote from Tom's post:
>Suppose the target statistic is "the mean weight of men attending >the upcoming commencement ceremonies at my school." I gather a >group of ten men and find their weights vary from 131 lbs to 245 >lbs with a mean of 162 lbs. My weights are accurate to within >+/- 0.5 lbs, so these are genuine fluctuations in weight that I >measure. Nevertheless, in terms of the target statistic, these >fluctuations are noise and contribute to the error bars of the >target statistic.
>Your fallacy, Surfer, is assuming that since I am measuring >genuine fluctuations in weight rather than experiencing random >measurement error, these fluctuations in weight do not contribute >to the error bars of the target statistic. In other words, you >believe that the mean weight of the target population of men as >estimated by my sample should be reported as 162 +/- 0.5 lbs.
>Imagine that one selected a different group of ten men. The >average for this second group is almost surely not 162 lbs. >Consider a third, fourth, fifth,... group of ten men, and plot >the distriution of the averages for the different groups. The >AVERAGES will display a variance, and that variance is related >to the variance of the weights of the individual men. THIS is >what statistics does: it tells you what the variance of the >average will be, given the variance of the individual >measurements (here mens' weights).
>For the case (like Miller's) where you have only one group of ten >men to consider, honesty precludes one from claiming the average >is 162 +- 0.5, and one must claim 162 +- sigma, where sigma is >determined from the distribution of the ten mens' weights. In the >language of statistics, the mean of those ten mens' weights is >the best unbiased predictor of the true average, and the sigma >is the best unbiased predictor of how accurately the average of >those ten weights reflects the true average. Note these are >"predictors", because one does not know the true values, one only >knows the ten values one measured.
> To learn how to compute that sigma you need to STUDY. If > those ten men's weights are randomly but uniformly > distributed between 131 and 245 lbs, the sigma (errorbar > on the average) will be about 10 lbs, not 0.5 lbs.
>That's PRECISELY what I did for each run of Miller's data: For >each of his eight orientations he averaged 40 data points. I >computed the variance of those eight averages from the variance >of the 40 points that went into computing each one. Those >variances (errorbars) GREATLY exceed the variation among the >eight averages, showing that the variation Miller used to make >his result is not significant. This, in turn, makes any >conclusion based on his results be insignificant: Miller >concluded the average is 11 km/s, but the errorbar on that >average is something like 100 km/s; Miller determined an average >direction, but the errorbar on that direction includes all >possible directions.
In addition to the false premises, I don't think a statistical error analysis of unprocessed partial values of a quantity can necessarily tell us anything useful about the final processed value.
Eg consider some equipment that can only separately measure hundreds, tens and units, so that after measuring two quantities it gives us the digits 5, 9, 8 and 6, 1, 2.
The final processed values of 598 and 612 differ by only 7.1%.
In contrast, the unprocessed partial values (the digits) differ by 10%, -80% and -60% of full range.
These huge variations tell us nothing useful about the variations in the final processed values. Or in other words, an error bar for the digits would tell us nothing useful about the errorbar for the final values.
In the case of the Miller experiments, a small change in direction of the velocity vector can produce a large change in the components measured by the interferometer.
Therefore the same kind of problem applies here. That is, an errorbar for the fringe shifts can tell us nothing useful about the errorbar for the velocity vector.
> >> It has nothing to do with what I want. > >> His paper contains false premises.
> >You are in broken record mode, Surfer. All of your points have > >either been previously answered, or are irrelevant "chaff" > >arguments thrown up to distract attention from the weaknesses > >of Miller's paper. So I am snipping.
> > [BIG SNIP]
> >You have NOT, however, adequately answered Tom or myself on the > >following. I challenge you to come up with a valid critique of > >Tom's procedure for determining errorbars.
> >If you cannot refute this procedure of elementary statistical > >analysis, then Miller's entire thesis falls apart.
> >I will begin with a "Director's Cut" quotation from myself, > >(replacing my blooper with my corrected remarks) followed by a > >quote from Tom's post:
> >Suppose the target statistic is "the mean weight of men attending > >the upcoming commencement ceremonies at my school." I gather a > >group of ten men and find their weights vary from 131 lbs to 245 > >lbs with a mean of 162 lbs. My weights are accurate to within > >+/- 0.5 lbs, so these are genuine fluctuations in weight that I > >measure. Nevertheless, in terms of the target statistic, these > >fluctuations are noise and contribute to the error bars of the > >target statistic.
> >Your fallacy, Surfer, is assuming that since I am measuring > >genuine fluctuations in weight rather than experiencing random > >measurement error, these fluctuations in weight do not contribute > >to the error bars of the target statistic. In other words, you > >believe that the mean weight of the target population of men as > >estimated by my sample should be reported as 162 +/- 0.5 lbs.
> >Imagine that one selected a different group of ten men. The > >average for this second group is almost surely not 162 lbs. > >Consider a third, fourth, fifth,... group of ten men, and plot > >the distriution of the averages for the different groups. The > >AVERAGES will display a variance, and that variance is related > >to the variance of the weights of the individual men. THIS is > >what statistics does: it tells you what the variance of the > >average will be, given the variance of the individual > >measurements (here mens' weights).
> >For the case (like Miller's) where you have only one group of ten > >men to consider, honesty precludes one from claiming the average > >is 162 +- 0.5, and one must claim 162 +- sigma, where sigma is > >determined from the distribution of the ten mens' weights. In the > >language of statistics, the mean of those ten mens' weights is > >the best unbiased predictor of the true average, and the sigma > >is the best unbiased predictor of how accurately the average of > >those ten weights reflects the true average. Note these are > >"predictors", because one does not know the true values, one only > >knows the ten values one measured.
> > To learn how to compute that sigma you need to STUDY. If > > those ten men's weights are randomly but uniformly > > distributed between 131 and 245 lbs, the sigma (errorbar > > on the average) will be about 10 lbs, not 0.5 lbs.
> >That's PRECISELY what I did for each run of Miller's data: For > >each of his eight orientations he averaged 40 data points. I > >computed the variance of those eight averages from the variance > >of the 40 points that went into computing each one. Those > >variances (errorbars) GREATLY exceed the variation among the > >eight averages, showing that the variation Miller used to make > >his result is not significant. This, in turn, makes any > >conclusion based on his results be insignificant: Miller > >concluded the average is 11 km/s, but the errorbar on that > >average is something like 100 km/s; Miller determined an average > >direction, but the errorbar on that direction includes all > >possible directions.
> In addition to the false premises, I don't think a statistical error > analysis of unprocessed partial values of a quantity can necessarily > tell us anything useful about the final processed value.
> Eg consider some equipment that can only separately measure hundreds, > tens and units, so that after measuring two quantities it gives us the > digits 5, 9, 8 and 6, 1, 2.
> The final processed values of 598 and 612 differ by only 7.1%.
> In contrast, the unprocessed partial values (the digits) differ by > 10%, -80% and -60% of full range.
> These huge variations tell us nothing useful about the variations in > the final processed values. Or in other words, an error bar for the > digits would tell us nothing useful about the errorbar for the final > values.
Your analogy is invalid, absurd and irrelevant.
> In the case of the Miller experiments, a small change in direction of > the velocity vector can produce a large change in the components > measured by the interferometer.
What craziness is this? Do you assert that the rules of vector math do not apply to the measurement of the components of a velocity along different axes?
> Therefore the same kind of problem applies here. That is, an errorbar > for the fringe shifts can tell us nothing useful about the errorbar > for the velocity vector.
Hypothetically, there is a direct relationship between fringe shift and the the velocity component in any given direction. Therefore an errorbar for the fringe shift along a given axis is directly proportional to the errorbar for the measured component of the velocity vector along that axis.
To suggest otherwise is to deny the validity of Miller's, Cahill's, and Allais' analyses, all of which assume a sensible relationship between fringe shift and velocity.
You are wounding yourself deeply in the foot, Surfer.
>> : His "work" in the field of physics is the "work" of a crank, very >> : close to the "qaulity" of Cahill's "work".
>> Nasa disagrees; please show his calculation errors.
> Both Allais and Cahill arrived to the same conclusion after:
> -not running any experiment > -reanalizing the Miller data
> ...that they could detect absolute motion. This is pure crank > material.
As I thought - please keep to the science instead of slander.
>> : At least Cahill is not an antisemite like Allais :-)
>> ??! That is a strong accusation, which makes you "anti-Allais" according >> to >> your use of the term - and it's just as irrelevant.
>> Harald
> Don't try to deflect the discussion, the point is about what Allais > writes on his website, it is pretty obvious that he's an antisemite. > This colors his "scientific" work.
It's prety obvious that you are a mud thrower; it's YOU who continues to deviate the scienctific discussions with unscientific remarks. Please stop.
> "harry" <harald.vanlintelButNotT...@epfl.ch> wrote in message > news:4837fa4b$1_6@news.bluewin.ch... > | > | "Androcles" <Headmas...@Hogwarts.physics> wrote in message > | news:RQRZj.103193$AN7.87552@newsfe23.ams2... > | > > | > "harry" <harald.vanlintelButNotT...@epfl.ch> wrote in message > | > news:4837e477$1_5@news.bluewin.ch... > | > | > | > | "Tom Roberts" <tjroberts...@sbcglobal.net> wrote in message > | > | news:99BZj.108$uE5.87@flpi144.ffdc.sbc.com... > | > | > Jerry wrote: > | > | [...] > | > | > Imagine that one selected a different group of ten men. The > average > | > for > | > | > this second group is almost surely not 162 lbs. Consider a third, > | > fourth, > | > | > fifth,... group of ten men, and plot the distriution of the > averages > | > for > | > | > the different groups. The AVERAGES will display a variance, and > that > | > | > variance is related to the variance of the weights of the > individual > | > men. > | > | > THIS is what statistics does: it tells you what the variance of > the > | > | > average will be, given the variance of the individual measurements > | > (here > | > | > mens' weights). > | > | > > | > | > For the case (like Miller's) where you have only one group of ten > men > | > to > | > | > consider, honesty precludes one from claiming the average is 162 > +- > | > 0.5, > | > | > and one must claim 162 +- sigma, where sigma is determined from > the > | > | > distribution of the ten mens' weights. In the language of > statistics, > | > the > | > | > mean of those ten mens' weights is the best unbiased predictor of > the > | > true > | > | > average, and the sigma is the best unbiased predictor of how > | > accurately > | > | > the average of those ten weights reflects the true average. Note > these > | > are > | > | > "predictors", because one does not know the true values, one only > | > knows > | > | > the ten values one measured. > | > | > > | > | > To learn how to compute that sigma you need to STUDY. If > | > | > those ten men's weights are randomly but uniformly > | > | > distributed between 131 and 245 lbs, the sigma (errorbar > | > | > on the average) will be about 10 lbs, not 0.5 lbs. > | > | > > | > | > That's PRECISELY what I did for each run of Miller's data: For > each > of > | > his > | > | > eight orientations he averaged 40 data points. I computed the > variance > | > of > | > | > those eight averages from the variance of the 40 points that went > into > | > | > computing each one. Those variances (errorbars) GREATLY exceed the > | > | > variation among the eight averages, showing that the variation > Miller > | > used > | > | > to make his result is not significant. This, in turn, makes any > | > conclusion > | > | > based on his results be insignificant: Miller concluded the > average > is > | > 11 > | > | > km/s, but the errorbar on that average is something like 100 km/s; > | > Miller > | > | > determined an average direction, but the errorbar on that > direction > | > | > includes all possible directions. > | > | > | > | Well explained this time! :-) > | > | > | > | Harald > | > | > Consider why did Einstein say > | > the speed of light from A to B is c-v, > | > the speed of light from B to A is c+v, > | > the "time" each way is the same, > | > | Easy: he did NOT say that.
> Yes he did, you stupid lying prat. > Quote: > "we establish by definition that the ``time'' required by light to travel > from A to B equals the ``time'' it requires to travel from B to A. " > Unquote. > Quote: > "But the ray moves relatively to the initial point of k, when measured in > the stationary system, with the velocity c-v". > Unquote.
Your above claim does not match your citation and an essential precision is lacking. Here's another one: "Walking speed is about 5 km/h" "The man has a speed, when measured in the stationary system, of about 1005 km/h". Contradiction?
> Where do you imagine the cuckoo malformations come from, thin aether?
"Thin" ether was apparently disproved by Lorentz. But a Lorentz type of ether as refined by Einstein is imaginable, so that everything is wave phenomena. What explanation do you have for the maximum speed of electrons?
|
