Question about Wald tests of linear hypotheses

[rebecca eagleton]

Hi everyone,

I'm conducting multi-group SEM in Stata 16.0. Because my data uses survey weights, I tested for group differences using Wald tests of linear hypotheses (as recommended elsewhere on the forum). Would someone be able to explain why this test is reported as an F statistic, rather than chi-square? Thanks for your help!

[Clyde Schechter]

Wald statistics reported as chi-square are calculated from estimators that are asymptotic. Basically you can think of a chi-square statistic as being like an F statistic with an infinite number of denominator df. This just like -regress-, calculating by OLS, reports a t-statistic (finite denominator df), whereas -glm- (maximum likelihood estimation) reports a z-statistic (normal distribution is like a t-distribution with infinite df).

[rebecca eagleton]

Hi Clyde,

Thank you so much for your answer. This is very helpful!