الفهرس | Only 14 pages are availabe for public view |
Abstract The goal of this research is to simultaneously estimate the parameters of the association-marginal (AM) model. This AM model is a special formulation, rather than special case, of the generalized log linear model (GLLM).This simultaneous AM model contains two models: the first model is an association (A) model which describes the association and interaction patterns among the responses and is specified as a standard log linear model with a log link. The second model is a marginal (M) model which describes the marginal distributions for the responses and is specified as a proportional odds model with a logit link. An advantage of simultaneous fitting rather than separate fitting is that greater efficiency of expected joint frequencies is obtained. Also, simultaneously fitting will provide a single test for overall goodness-of-fit and it provides improved model parsimony, resulting in better estimates than one would obtain by fitting the models separately. |