Re: Dayton Miller's Data have no Real Signal

demeo@xxxxxxxx wrote:
Tom Roberts wrote:
Title: Dayton Miller's Data have no Real Signal
Author: Tom Roberts
Date: December 8, 2005
[...] 

Firstly, you can review my papers on the Miller ether-drift subject
here:
http://www.orgonelab.org/miller.htm
http://www.orgonelab.org/DynamicEther.pdf

You make the same mistakes Miller did, and the same mistakes everyone else who claims his result is valid: you did not include errorbars. Yes, Miller did not provide errorbars, but he was writing before their usage became commonplace (or rather, compulsory). But his data contain voluminous information about them. For plots such as Figure 1 of http://www.orgonelab.org/miller.htm, the errorbars exceed the size of the paper. That is, the "result" you claim is not _SIGNIFICANT_.


I note you cite Miller's 1933 paper, but nothing else aside from the
hit-piece authored by Shankland et al.  There are dozens of papers
which bear upon these issues, [...]

Sure -- this is a newsgroup article, not a journal paper. So far. I want comments, precisely like yours, before preparing a real paper.


You evaluated Miller's August 1927 data set, but [...]

I evaluated all the data I could get my hands on easily, and for which I had a reasonable provenance (I obtained it courtesy of Glen Deen). MSN groups is not letting me log in, but when I looked at the group DaytonMillerData several months ago, the files had no sensible provenance, and I chose not to use them. If you have computer-readable files with provenance, LET'S TALK (via email: tjroberts@xxxxxxxxxx ).


Shankland, et al, did their best to bury Miller's work forever.  They
failed, [...]


I agree, but for very different reasons....

No matter, the DSP approach I have been applying can settle the matter. For the data I have, there is no doubt that there is no real signal. I am interested in applying it to other data of his if I can obtain it without undue effort. But I think you are incredibly naive to think that his Cleveland data could nave no signal at all, and yet his Mt. Wilson data have a signal. I think the only way to know is to apply a modern data AND ERROR analysis to that data.


[my critique has not] addressed the
SYSTEMATIC SIDEREAL-DAY VARIATION IN THE AXIS OF ETHER-DRIFT, APPARENT
DURING ALL FOUR SEASONAL EPOCHS. 

You can find faces in clouds, if you look hard enough.

For very noisy data, such as Miller's, there is indeed a best-fit value, and in many cases simple techniques can find it, and Miller apparently did. But for all of the data I have, and for the plots in Miller's 1933 article, it is QUITE CLEAR that the answer obtained is not SIGNIFICANT. That is, one can fit the data in those plots equally well using ANY direction whatsoever, and ANY speed less than some value >30 km/s. This is not a "determination of the earth's absolute motion" by any MODERN standard.

In 1933 Miller could get his paper published; today he would not, for the simple reason that it lacks an error analysis.


The pattern was systematic, 

But not SIGNIFICANT.


The Cleveland experiments, pre-Mt.Wilson or post-Mt.Wilson, are not
significant for the question of ether-drift on a number of counts.
Firstly, as mentioned, they were mostly calibration experiments -- can
you cite a publication somewhere which reports on those post-Mt.Wilson
experiments in Cleveland? Are you sure you do not have, by some
accident, the data sheets from 1927 of Michelson-Pease-Pearson? ( If
so, I'd like to get a copy of them!) I would ask, where did you get
your set of Miller data sheets?

I obtained Excel files from Glen Deen. Their provenance is in their contents, which state (among other things): Cleveland; August 1927 and Seoptember/October 1929, plus I transcribed myself the Fig 8 data from September 1925.

This is an awful lot of "calibration experiments", over an awfully long time.


The only ones I know about include the
full sets from Mt. Wilson, so it seems strange to me that you'd get
only the 1927 data sheets, but not the others from 1925 and 1926

I only have the data Glen transcribed. I don't know what criteria he used to select pages for transcription.

I am aware this is a problem, and would love to get more data. But I don't have the resources to transcribe it myself.


Sorry to
say, Tom, your analysis is faulty on a number of levels, and does not
touch Miller's findings and conclusions 

Why so?


 It is a pity you did not consult with the advocates
of ether-drift prior to undertaking your analysis,

Send me data files, and the analysis will take only a short time. My biggest problem is finding time to devote to this.

AFAIK none of the "ether advocates" have done ANY realistic or modern data and error analysis of these data -- they just marvel at the supposed sidereal variations, without ever asking "are they SIGNIFICANT?". And then they usually foam at the mouth about how some vague "science establishment" has "suppressed" Miller's wonderful result. That is wishing and hoping, not science.


Any _MODERN_ analysis of Miller's data will perform an error analysis, propose a theory or model to which the real signal can be compared, and then display via plots WITH ERRORBARS how well (or poorly) the data agree with the model. If an "absolute direction" or "absolute speed" is determined, those values will NECESSARILY have errorbars attached. Your papers do none of that, and neither do Allais', or Consoli's, or Cahill's, or anybody else's I have found. Except mine (which is currently just in early draft form -- I haven't gotten to the model stage yet, but it's clear SR is consistent with the data I have).


But I still don't see how your method
can do more than point out the obvious, that the signal is often buried
in the noise.

If a quantitative model of the noise [#] completely describes the data, then there is no room for any real signal. That is the case for all of the data I have.

	[#] Which I call "systematic error" -- "noise" has IMHO the
	    wrong connotations. What I would call the noise in
	    Miller's data is quite small, <0.2 fringe; it is due
	    primarily to roundoff errors by the observer.


I suggest firstly to review my
papers on the subject, so I won't necessarily be repeating myself.

I already have. You make the same mistakes everyone else did -- lack of error analysis, and lack of a MODERN analysis of the data (e.g. why haven't you reported results of DFTs?).


Have you ever even LOOKED at Miller's data?? -- please spend the time to actually plot the RAW data from Miller's Figure 8 and LOOK at it! (be sure to add back his adjustments and plot all 321 data points) -- Then say with a straight face that you still think there is actually a "signal" in there.... Or even that Miller's subtraction of a LINEAR error is even remotely valid. Until you do at least this plot you don't understand his data at all, no matter how much you have "studied" it. And if you have already made such a plot, why doesn't it appear in any of your papers?

The overall appearance of this plot casts grave doubts on any claim to a real signal in there, and CERTAINLY says that without a full and complete error analysis no such claim can be believable.

	[FWIW that plot shows an extremely ragged sequence, fluctuating
	 about +- 1 fringe around a line with slope -.3 fringe/turn;
	 the fluctuations have irregular periods > 1 turn. The supposed
	 "signal" (with period 1/2 turn) is invisible to the naked eye.]


Tom Roberts	tjroberts@xxxxxxxxxx
.