Horace Heffner wrote:
On Jan 21, 2006, at 1:51 PM, Taylor J. Smith wrote:
Source: David de Hilster ([EMAIL PROTECTED])
``Dr. Domina Eberle Spencer still has the raw data from
the Hafele- Keating experiment (atomic clocks on planes
in 1972) which she examined in 1996 and concluded that
that data to prove Einstein was fabricated. She is a
brilliant mathematician and has agreed to be interviewed
on the subject for our documentary film.
That should be an interesting film, but I doubt anything substantive
would be put on film. I hope she publishes. Without the critical
data, it is not possible to reach any conclusions. In fact, even with
the data, the only conclusion might be there is a need for more data.
Since last posting to this thread, I've read one paper deconstructing
the H-K experiment and read a bit more about the topic in general.
It appears to me that the H-K experiment was indeed BOGUS. It wasn't
supposed to be that way -- the intent of the researchers was apparently
pure -- but in the end it suffered from "Not Enough Rats" syndrome.
This problem is common in biology and the social sciences but unusual in
a physics experiment. Here's a sketch of how it happens in a biology lab:
A researcher whom we'll call Bob wants to determine the effect of diet
on the neurotransmitter Poodlecatamousitine. He understands statistics
well enough to analyze his results with no difficulty, but is none the
less a little shaky on the use of statistics during the experimental
design phase (this is all too common, don't say it doesn't happen!).
But he _guesses_ that 20 rats in each of his two experimental groups and
20 more in the control group should produce a clear enough result.
But rats are expensive, grant money's tight, and he decides he can make
do with just 10 rats in each group.
But three of the rats get Rat Flue and check out before the end of the
experiment, there's an air conditioner failure and two more shuffle off
this mortal coil, ALF raiders get several more, and one of the "male"
rats gets pregnant and is disqualified. Two others refuse to eat the
special diet and are also disqualified.
They get to the end of the experiment, and it's time to gather data.
But the only way to get good readings on Poodlecatamousitine levels is
to sacrifice the rats in total darkness, by beheading them while they
sleep, and then take brain slices (don't laugh; I knew someone who had
to do exactly this in a rat experiment). Unfortunately the grad
students find this difficult to carry out. One gets bitten and yells,
waking up several rats; they're disqualified. Two rats get mixed up in
the dark, and so they're out, too. And one of the "brain slices" turns
out to contain almost nothing but human finger tissue, and so it's no good.
In the end Bob has data for 4 test rats in group A, 5 in group B, and
just three control rats. The "large effect" he hoped to obtain turns
out to be a difference of just 2% in the levels, and it's ... **Not
Statistically Significant**. Oops. But this represents months of work,
so maybe Bob publishes anyway.
Back to the H-K experiment...
H-K flew 4 clocks around the world, but it appears that they should have
used several times that number. It seems that portable cesium clocks
are not the paragons of reliability I had thought they were and the
clock drift rates -- and rates of change in the drift rates! -- were
erratic and large compared with the effect being measured. All but one
of the clocks they used showed large (and obvious) changes in drift
rates, and the last clock, which didn't _obviously_ have a large change
in drift rate (but may none the less have changed drift rate more than
once during the flights), didn't show the Sagnac effect. I don't know
if it's possible to do a good stat test on their results but the
discussion I read made it sound like their results would not have gotten
over the bar had they been properly analyzed.
H-K apparently chose to publish anyway, and in fact, they supposedly
didn't release their raw data until a very long time after the paper was
published. That's always a red flag, of course -- if there's nothing to
hide, there's no reason not to release the raw data. In this case it
appears to me that the experiment really proved nothing either way
except to show that, if we want to test this effect using that method,
either better clocks or more clocks are needed, or they need to use much
faster aircraft (the magnitude of the predicted effect doesn't change if
you go faster, but the noise level would be reduced).