|
Abstract:
|
In clinical trial and many other applications proportional odds model are fitted to model ordered categorical data. In real world application, especially in clinical trial data, presences of missing data are inevitable. Missing values in the data occur due to different missing mechanisms, such as missing completely at random (MCAR) and missing at random (MAR). In the regression set up with missing data, most of the focuses are on missing covariates; however, very little attention has been paid to missing responses. If the responses are missing, and the missingness depends on the responses itself, then it is called nonignorable missing. In our work, we will focus on how to handle missing responses in fitting proportional odds model when the missing data are nonignorable. Following Ibrahim and Lipsitz (Biometrics 52: 1071-1078, 1996) we will propose an EM algorithm to fit the model. We will investigate the bias correction of the estimates of the regression coefficients. All of these novel methods will be demonstrated using real world data.
|
