Does Bivariate Normal have a monotone likelihood ratio?

165 Views Asked by At

In general, with all parameters unknown I think the answer to this question is no. I think this because in this instance we would have a curved multivariate exponential family. Is this reasoning correct?

Now, suppose the only unknown parameters are the correlation, $\rho$, and one of the means, $\mu_1$, then in this (or any other) particular instance does the bivariate normal distribution have a monotone likelihood ratio?