|
Abstract:
|
The proportional odds model is widely used for ordinal categorical responses. Since the likelihood equations are intractable and do not have explicit solutions, maximum likelihood estimation of the model parameters requires iterative procedures. We propose using the modified maximum likelihood (MML) method to lessen the difficulties encountered in maximum likelihood estimation of the proportional odds model. We show that the MML estimators, which are explicit functions of sample observations, are asymptotically fully efficient under some very general regularity conditions and highly efficient for small sample sizes. We study the robustness properties of these estimators via simulations. We also illustrate the method using real life examples.
|
