تحليل پاسخ هاي دسته اي رتبه اي طولي با مقادير گمشده به روش بيزي

The longitudinal studies, in which response variable is ordered categorical data, such as pain intensity, degree of recovery and nausea intensity and etc, have special setting in medical sciences, because physician examines the patient frequently during the period of study. One of the most important problems in analysis of such studies is missing data. The model which is used in these states, is the Generalized Linear Model shown F(n) = x,u . In this article by considering correlation between observations of each subject, Dʹs coefficient in two states of complete cases and existence of missing values in the practical example were estimated and compared with each other by using Bayesian approach in SAS and Win bugs soft wares. Results showing that analysis through Bayesian approach and with introduced model, Significant variables have not changed in the state of complete cases and existence 10% or 20% missing values, while by omitting the data of 20% subjects having missing values, no variable has been significant.Omitting the data related to the cases that have missing data, causes theaccuracy of analysis to be reduced. Therefore, using Bayesian approach (even with Non-Informative Priors) in longitudinal studies with missing values is recommended with the introduced model.