The sub-model with zero rat variance can be fitted by omitting the relevant term from the model formula in either syntax. The conventional log likelihood ratio statistic is twice the increase in log likelihood for the larger model relative to the submodel. Check that these functions report different numbers for the log likelihood, but they return the same log likelihood ratio. Report the value. (Recall that the log  likelihood is defined up to an arbitrary additive constant, which may depend on the response or the design matrix.)