lrtest

[Guest]

Hello,

I have three variables and want to see if it is worth including all of them in the model or if the lrtest indicates it would be better to just include one or two variables. The three variables are all categorical. Can anyone advise how I enter this?

 clogit model 1
estimates store m1

clogit model 2
estimates store m2
lrtest m1 m2

For the first line model 1 do I enter all three variables? If so, what is entered in model2? Or do I create three models with each variable added individually?

[Marcos Almeida]

By using the - lrtest -, we abide by the assumption that the second model is nested in the first.

And Stata is telling you that.

For example, if we type:

. lrtest m2 m3

Likelihood-ratio test                                 LR chi2(1)  =      6.64
(Assumption: m3 nested in m2)                         Prob > chi2 =    0.0100

Hopefully that was helpful