Talk:Score test

From Wikipedia, the free encyclopedia

Shouldn't "identifying the C above with log(K)" actually read "identifying the C above with log(K)/n"?

Anais: And what happens under the alternative hypothesis (if H0 is not true) ? Does it tend to infinity ? It is quite important if you want to define the critical region... —Preceding unsigned comment added by 90.46.131.123 (talk) 11:26, 20 January 2008 (UTC)

The initial single-parameter section and the "multiple parameters" section take different approaches at present that need to relate to each other better. The single-parameter section is showing that a test based on the score is locally most powerful (LMP) without saying anything about its distribution or mentioning Fisher's information I. The multiple parameters section gives the asymptotic distribution of the score test statistic, which involves both the score U and the information I, without saying it's LMP. Would be better to do the whole thing for a single parameter first then show how it generalises to multiple parameters. Afraid i don't have the textbooks to hand that i'd need to be sure about changing it myself. --Qwfp (talk) 17:52, 23 January 2008 (UTC)