贝叶斯对数线性估计

Bayesian log Linear Estimates for High-Way Contingency Tables

用Logits变换方法来讨论列联表的贝叶期估计，这时的估计称为贝叶斯对数线性估计，最早是由Lconard（1973，1975）提出的．Leonard讨论了二维表的对数线性贝叶斯估计；其后，Nazerat（1987）给出了三维表的贝叶斯对数线性估计．该文采用矩阵求微商方法，立足于高维表贝叶斯线性估计的较一般的表达形式；具体对四维表讨论员叶斯对数线性估计的表达和计算．

This paper presents a Bayes method to estimate the expected proportions for high-way contingency tables appropriate when prior knowledge about the interaction effects can be described by a particular kind of exchangeability assumption.We give the expressions of estimates for four-way contingency tables based on general expressions equations.It extends the work of Leonard (1975)and Nazaret(1987) and useful algorithm also discussed.