The result of "test" do not align with the "LR test"

[Howard Wu]

Hi all,

I am running logistic regression. Model 1 is the nested model. Model 2 is the comparison model with interaction terms.

*race have four categories
*x1 two categories
*x2 two categories

Model 1:
logit y i.sex c.age i.race i.ses i.x1 i.x2
estimate store model1

Model 2:
logit y i.sex c.age i.ses i.race##i.x1 i.race##i.x2
estimate store model2

testparm i.race#i.x1 i.race#i.x2
lrtest model2 model1

Wald test: chi2 (6) = 11.47, p = .0749
LR test: LR chi2 (6) = 11.57, p = .0723

Why there is a difference here? Thanks!

[Maarten Buis]

The Wald test and likelihood ratio test only asymptotically give the same results. In finite samples the results will differ. Anyhow, the results seem similar enough, so I don't see a reason to be worried.

[Howard Wu]

Thank you, Maarten!