Path: ...!weretis.net!feeder8.news.weretis.net!eternal-september.org!reader01.eternal-september.org!.POSTED!not-for-mail From: Anton Shepelev Newsgroups: sci.physics.relativity,sci.stat.math Subject: Re: statistics in Roberts' paper on Miller Date: Sun, 12 Mar 2023 01:51:41 +0300 Organization: A noiseless patient Spider Lines: 109 Message-ID: <20230312015141.f3847802df5f63148bd9b8f1@gmail.moc> References: <20230219220058.8d3d14741e18cce1bf19e256@gmail.com> <51151e80-a719-46ef-8095-6535309e7d02n@googlegroups.com> <20230220003936.ca90df6f8848a095271a0cbe@gmail.com> <20230223193132.41882edd1d9110b60e745dac@gmail.moc> <20230225001353.60271597ed5a42bec16e8d54@gmail.moc> <0u3qvhlnu50kk3kg7e7jn6ujnene2fo8jk@4ax.com> <6b425793-9cd8-4da4-bdf5-1e245b9017a3n@googlegroups.com> <20230311011444.39673053ab2d3d20d4614e6d@gmail.moc> MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Injection-Info: dont-email.me; posting-host="55af21dbd8cd739ec117f18769a4a0a2"; logging-data="2865844"; mail-complaints-to="abuse@eternal-september.org"; posting-account="U2FsdGVkX18A3GTRlT76r6s7STXVpnjJkHVKjJH59aE=" Cancel-Lock: sha1:fAH1DJaddNIkvH643UdE1BgDpZg= X-Newsreader: Sylpheed 3.7.0 (GTK+ 2.24.30; i686-pc-mingw32) Bytes: 5968 Tom Roberts: > Anton Shepelev: > > RichD: > > > Tom Roberts: > > > > > > > Worse than lack of statistical errorbars is Miller's > > > > lack of knowledge of digital signal > > > > processing -- his analysis is essentially a comb > > > > filter that concentrates his systematic error into > > > > the DFT bin corresponding to a real signal -- that's > > > > a disaster, and explains why his data reduction > > > > yields data that look like a sinusoid with period > > > > 1/2 turn. > > > > > > Can you elaborate on this filter? > > > > Mr. Roberts is referring to the procedure of "folding" > > the data of each 16-azimuth turn into an 8-azimuth half- > > turn by summing up the observations at azimuths 180 > > degrees apart. > > No. I am referring to Miller's averaging the 20 turns. Sorry for the misunderstanding, them. Yes, Miller averaged his observrations over 20 or more (full) turns. > As his final result for a single run is the plot at the > bottom of my Fig. 1, with 8 points, No, that plot is not the final result of his analysis, but a digression made for the purposes of merely a "preliminary study" -- quoth Miller: For the purpose of a preliminary study of the observations, it is convenient to obtain an approximate graphic representation of the effect by the following procedure. The second half of the line of sixteen average readings is placed under the first half and the mean of the two numbers in each column is obtained; These half-period plots were /not/ used in the final analysis (see below). > my discussion is of the fact that he averaged 40 values to > get each point of the plot. Yes, he did average the 40 values for that plot, that is averaged the 20 turns /and/ then folded the result in two. But no, he did not employ the last (half-turn) averaging in his actual data analysis: In the definitive study of the ether-drift effect, this set of sixteen average readings for the position of the interference fringes is plotted to a large scale and is subjected to mechanical harmonic analysis to evaluate precisely the second harmonic component, which represents the second-order, half-period ether- drift effect; [...] The twenty or more readings for each of the sixteen observed azimuths are averaged and the averages are compensated for the slow linear shift of the whole interference system during the period of observation, as explained previously in connection with Fig. 9. The average readings for each set are then plotted on coordinate paper, to a large scale, for the purpose of harmonic analysis. [...] These charted "curves" of the actual observa-tions contain not only the second-order, half- period ether- drift effect, but also a first-order, full-period effect, any possible effects of higher orders, together with all instrumental and accidental errors of observation. [...] In order to evaluate precisely the ether-drift effect, each curve of observations has been analyzed with the Henrici harmonic analyzer for the first five terms of the Fourier series. The quotations above provide compelling evidence that Miller did not combine the half-turn observations in his analysis. Your second comb-filter, which raises the lowest DFT bin up to the half-turn frequency, is absent from Miller's procedure, whereas your first comb-filter, which raises the lowest DFT bin up to the fundamental full-turn frequency, is no doubt present. > But this still holds for his averaging of all 16 > orientations -- it is still a comb filter, and with a > rapidly-falling noise spectrum it pushes most of the noise > into the lowest DFT bin. Yes, no good, but in that case the lowest DFT bin is the fundamental, full-perdoid, full-turn frequency, not the half-turn one. > Averaging raw data is a VERY BAD analysis technique. But > back in 1933 this was not understood; we understand it > today. Absolutely correct, even as any transoformation that reduces the amount information in the data before that information can be availed of. -- () ascii ribbon campaign -- against html e-mail /\ www.asciiribbon.org -- against proprietary attachments