How can I test any significant difference between two models


New Member
I'm fitting a poisson regression model but using two different likelihood functions to estimate the parameters.

I then want to check that model is more significant than the other.

I was reading and I found one paper suggesting that I should use a hausman test, but this does not make much sense to me. Could anybody suggest any other method/test or just a way that I would show this?


Omega Contributor
Identical model terms used? In multilevel modeling where everything is the same accept, say which covariance structure you use, it is common to compare things like BIC or AICC. You don't necessarily need to conduct null hypothesis significant testing, since they are the same models and you are just looking at fit parameters.


Fortran must die
If the model is nested within each other there is commonly a test if one is better. For example in multilevel modeling or SEM such test exist and are commonly used. I do not know if that is true for poisson models which I rarely use.