Re: Miller's errorbars
- From: Tom Roberts <tjroberts137@xxxxxxxxxxxxx>
- Date: Fri, 4 Sep 2009 07:47:58 +0000 (UTC)
Surfer wrote:
On Thu, 27 Aug 2009 07:05:14 +0000 (UTC), Tom Roberts
<tjroberts137@xxxxxxxxxxxxx> wrote:
I don't deny that there is a nonzero amplitude forConsider a radio antenna with 1 volt peak to peak of noise and 1
a Fourier component with period 1/2 turn. But:
A) It is not statistically significant (the errorbars are MUCH
larger than its amplitude)
microvolt of signal.
The errorbars for the raw data from the antenna would be a million
times larger than the signal.
Would you then say that the signal is not statistically significant?
I don't think so, because I think even you would accept that it is
quite legitimate to use a tuning circuit to select a radio signal and
reject the noise.
Your analogy is completely irrelevant, because a radio receiver makes
exactly 1 measurement of the input at each instant in time, and applies
a tuned filter to extract the narrowband signal of interest. That is,
the radio does not make repeated measurements of the signal at a given
time (obviously it cannot do so).
Miller, on the other hand, made 40 measurements of THE SAME VALUE
(fringe position at a given orientation), for 8 different orientations.
The problem is that those 40 different measurements of THE SAME VALUE
differ from each other by an amount 100 times larger than the amplitude
of the "signal".
The radio cannot re-measure the signal at a given time;
Miller, however, CAN AND DID re-measure the signal at a
given orientation, 40 times per run. This repetition is
what makes it possible to RIGOROUSLY determine the
errorbars for Miller's result; this is not possible for
the radio.
A better analogy is an attempt by a VERY determined experimenter to
measure the voltage of eight 1.5 V batteries using a faulty voltmeter --
for 40 measurements of each battery, the meter reads values ranging from
-75 to +75 volts. You could not reasonably suppose that this is
measuring the batteries at all, but is rather exploring the
instabilities INSIDE the device that was labeled "voltmeter" but clearly
DOES NOT ACTUALLY MEASURE THE VOLTAGE BETWEEN ITS LEADS.
[Miller considered his interferometer to measure "aether
drift". It is QUITE CLEAR he is really exploring the
instabilities of the instrument; this is probably the
most unwieldy thermometer ever constructed.]
That is Miller's basic problem: his instrument has a combined drift and
noise level that is ~100 times larger than the "signal" he claims to see
(for this run). The NONLINEAR nature of the drift and the ENORMOUS level
of noise are two different criteria, and each one completely invalidates
his claims.
No amount of rationalization on your part can change that.
If you had watched the hapless experimenter with that
faulty voltmeter, you would have no problem dismissing
the results. So do that for Miller -- LOOK at Fig 3 of
my paper (which simply plots Miller's data for this run
at a single orientation). Every corner and dot is from
a single orientation, so in the analogy they correspond
to a specific battery; CLEARLY his instrument is faulty.
(My analogy above corresponds quantitatively to Miller's
data for this run -- the "signal" he found is .06 fringe,
and those corners and dots vary by 6 fringes. So for a
1.5 volt signal I made the values vary by 150 volts.
Moreover, the OUTRAGEOUSNESS of ignoring enormous
variations is also comparable -- there are many
measurements of the anisotropy in the speed of light that
have vastly better resolution than Miller's "signal",
but see isotropy.)
Tom Roberts
.
- Prev by Date: Re: String theory and NCG?
- Next by Date: Re: Bertlmann's socks and Einstein's other shoe
- Previous by thread: Re: String theory and NCG?
- Next by thread: Re: Miller's errorbars
- Index(es):
Relevant Pages
|
Loading
