Re: Request for feedback on statistical method
From: George Kahrimanis (anakreon_at_hol.gr)
Date: 11/06/04
- Next message: R. Martin: "Re: Why were the exit polls wrong?"
- Previous message: George Kahrimanis: "Re: Why were the exit polls wrong?"
- In reply to: Shepherd Moon: "Request for feedback on statistical method"
- Next in thread: Anon.: "Re: Request for feedback on statistical method"
- Reply: Anon.: "Re: Request for feedback on statistical method"
- Messages sorted by: [ date ] [ thread ]
Date: 6 Nov 2004 07:58:00 -0800
Hello. My second degree is in phyics, but I have not yet had time
to browse the articles mentioned in the previous postings in
this thread. I remember from school that (the great) Dirac once
introdueed variability of the speed of light in a toy theory,
which is not taken seriously because it conflicts with
geological data.
If the hypothesis of slowing speed of light is to be taken
seriously, predictions ought to be made and a new measurement
should be performend, to see which prediction is right. Short of
that, we are talking nonsense.
Before I read the papers (perhaps I will not) I must
highlight the shortcomings of founding inference on confidence
intervals. The coverage probability, a.k.a. "confidence", applies
while one is designing a procedure, for the outcomes that *will*
be gathered. It does not imply probability after the data are
known, as many eager scientists want to interpret it. I shall be
glad to explain this point further and cite references.
To provide a graphic description, imagine that the designer of a
procedure programmes some robots to perform that procedure; he
has confidence 99%, say, that each intermediate conclusion will be
correct, but his condidence does not extend to us, the robots on
the field. After we see an outcome, the 99% probability is voided
for us, and the (otherwise impeccable) theory of confidence
intervals has no provision for us poor robots. We are just to do
or die.
Since the claim of slowing speed of light is expressed in the form
of confidence intervals, I have nothing to say, really. I suppose
that if I read the papers I shall find some mistakes, but this is
not a serious pursuit, imo.
This issue reminds me of that study from Germany, claiming that
it is good for a student to repeat the classes of a whole year.
(Again, thanks to Rich Ulrich for pointing out the mistakes in
my first review of that analysis.) It seems that statistics is
the last refuge of every outrageous hypothesis. I could go on but
this message is too long already.
~ George Kahrimanis
- Next message: R. Martin: "Re: Why were the exit polls wrong?"
- Previous message: George Kahrimanis: "Re: Why were the exit polls wrong?"
- In reply to: Shepherd Moon: "Request for feedback on statistical method"
- Next in thread: Anon.: "Re: Request for feedback on statistical method"
- Reply: Anon.: "Re: Request for feedback on statistical method"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
