Hi everyone

I'm just create a model selection table with MuMin Package in R for a GLM. To do this, I specify the global model which includes all independent variables (X) and the response variable (Y), and tested the goodness of fit using Chi square test (pchisq in R). To my first global model, I used a Poisson distribution, but the Chi square test rejected the null hypothesis (Overdispersion). So, I decide to use negative binomial distribution, and for this GLM, the test accept the null hypothesis.

After this, I use MuMIn to create a table model selection which create a subset of models derived of the global model, and testing each independent variable like in stepwise process. But for this package, i can´t calculate the Chi square for the models subset. Should I still care about the goodness of fit of my subset of model after knowing that for my global model it fits?