Re: VMs: Gordon Rugg's study follow ups

[Jacques Guy <jguy@xxxxxxxxxxxxxxxx>]

4/12/2004 1:14:40 AM, Dennis <tsalagi@xxxxxxxx> wrote:

>	Is it indeed true that phi is a measure of
>the magnitude of an observed difference, in a way that
>chi2 is not?

Yes. Chi2 is a measure of the _certainty_ that there
is a difference, however tiny. Take a roulette wheel
for instance. If over a million spins red wins 510,000
times and black 490,000 times, then you can be pretty
certain that the roulette cylinder is warped, but 
that's only 51% vs 49%, not enough to justify playing
red, since the casino's take is just under 3% (in the
case of a roulette with only one zero). Chi2 will tell
you that it is 99.99% certain that the roulette is
skewed. Phi will tell you how skewed it is.

>	I wonder what chi2 you would see between two texts in the
>same language by the same author but about different
>subjects.

Or between two authors on the same subject.

In the first case (same author, different subjects)
how certain it is that the subjects are different.
In the latter the authors. The longer the texts,
the likelier chi2 is to see a significant difference.


> As I recall, the stylometric methods used in
>the New Testament and Book of Mormon studies I mentioned
>compared the frequencies of common words like "the", "and",
>"where", "before", and the like.  If so, these
>methods might well show the same chi2 for the two texts.

Look at what you just wrote! "The same chi2 for the two
texts". You don't have a chi2 for text A and another for
text B. But you have it for text A compared to something
else. Text B for instance. Or anything else.



______________________________________________________________________
To unsubscribe, send mail to majordomo@xxxxxxxxxxx with a body saying:
unsubscribe vms-list