T-test in the proc GLM

Mathis1 — Tue, 21 Apr 2020 16:11:23 GMT

Hello,

I ran a proc GLM to predict a model with a countinuous variable et some categorical variables.

For each modality i have in the output the coefficients and also the t-test p-value. I'd like to know how the t-test stat is calculated. I'm not sure how it works for categorical variables. Actually, i'd like to know what "u" is used for the calculus of the test : ( mean(X) - u)/ (sigma / root(n)) .

In other words, what difference do we try to interpret, when looking at interpreting the t-test for a modality of a categorical variable ?

Thank you by advance 😉

Re: T-test in the proc GLM

PaigeMiller — Tue, 21 Apr 2020 16:55:17 GMT

u is zero, the test is if the coefficient is zero, or in other words (and in layman's terms), it is a test if the coefficient has an impact or is essentially zero and has no impact

If you want to test (for example) whether or not the levels are statistically different, that would be the F-test for the variable.

topic Re: T-test in the proc GLM in Statistical Procedures

T-test in the proc GLM

Re: T-test in the proc GLM