> I didn't forget the word "test". What
> you do with the number is distinct from
> what you call it. To me, every test needs
> a statistic. No matter.
The goodness of fit measures I mentioned (with the cutoffs 0.9 and
1.0) are derived from the likelihood ratio test statistic. So what
practicioners chose to do is to forget about its distribution and use
it as a summary of fit, rather than do the hypothesis testing with it.
That is to say, the test statistics tend to become omnibus measures.
Also, I've seen people deriving F-statistics for testing an increase
in R2 as you add variables to the model. So it is the R2 that is the
"first principle" for them. The distinction between "measures" and
"test statistics" may become quite blurred when it goes into the
field. You can have the distinction just fine in your head, but not
everybody wears that one on their shoulders :)
> Mathematical statisticians have no rituals??
>
> "Let ... be iid ..."
>
> "Assume a stationary process...."
>
> If mud is to be slung, let it slung
> in all directions.
That's right now! As a Russian saying goes, "One can see a speck in
somebody else's eye, but fails to notice a log in his own one".
--
Stas Kolenikov
http://stas.kolenikov.name
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
