I am trying to conduct a joint hypothesis test on parameters from a pooled regression (with PCSE and AR1 correction). One regression is restricted to have fewer variables (i.e. some Betas set = 0). The log likelihood function for the the restricted model is larger than the restricted model. The unrestricted model has 2 vectors of explanatory variables X1 and X2 and the restriced model just has vector X1. Shouldn't the log likelihood of the unrestricted model be larger? What am I missing here?