the criteria of choice

[zaineb belkhiria]

Hello
I hope you are doing well.
I have a question, please, about the criteria for choosing between the probit and xtprobit models.
Other than the test "log likelihood ratio of rho", is there any other method that allows me to choose between the probit and xtprobit models?
I hope you can help me, and I thank you in advance.
Best regards.

[Maarten Buis]

Why not just look at rho and make an informed decision on whether you think it is high enough to warrant a model or small enough that simpler model suffices? Remember that the definition of a model is that it is a simplification of reality, and simplification is just another word for "wrong in some useful way". So, by definition there is a lot of wriggle room in deciding what is a good model and what is not a good model. So any procedure that claims to come up with a yes or no answer does not accurately capture the nature of the problem; it is hiding part of the problem and that is bad. The key to making model choice "scientific" is not to follow protocol xyz, or do test abc, but to be open and transparent about what you have done. The protocols and tests are there to help, they play a supporting role, but you as the researcher play the main role. They make suggestion, but you make decisions and you can and should overrule the suggestions if you think the suggestions are wrong for your particular problem.

[zaineb belkhiria]

Hello.
I have already taken into consideration the value of Rho and the results of the test "log likelihood ratio of Rho" to confirm that the xtprobit model is the best model to make the estimation. I thought that there were other methods to make the right choice between probit and xtprobit.
Thank you, sir.