I'm trying to use Vuong's test to compare LR models - having trouble with the definition of 'strictly nonnested' vs 'partially nonnnested'/overlapping.

There are a few (lay language) definitions of ‘strictly non-nested’ floating around that suggest that models must be quite different to employ Vuong's [e.g...strictly nonnested if they have "different distributional assumptions on the errors, say normally or logistic distributed" or if they have "the same distributional assumption but different functional forms" such as if one is linear and the other is nonlinear”], but at the same time I see it used for testing models with "just" different parameters, i.e.
M1 <- glm(B1 + B2, family=binomial)

M2 <- glm(B3 + B4, family = binomial)

Vuongtest (M1, M2)

Any advice or reference appreciated. I myself cannot understand Vuong's paper but I did find this that looks like it contains a formal definition.

Can anyone comment?

enter image description here

  • $\begingroup$ I have trouble with your " LR models". $\endgroup$ – user158565 Dec 5 '18 at 4:05
  • $\begingroup$ Sorry, I mean 'logistic regression' models although for the purposes of defining fully vs partially non-nested I dont think it matters. $\endgroup$ – Marina_ANOVA Dec 7 '18 at 21:59
  • $\begingroup$ I thought LR = likelihood ratio. $\endgroup$ – user158565 Dec 7 '18 at 22:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.