Proportional Odds Model Affiliate with Random-Effect Longitudinal Model

Joint survival-longitudinal analysis gains popularity in recent clinical studies. A proportional hazards (PH) model in survival sub-model is commonly an alternative path to simplify a complex covariates hazard model into a regression model. The PH model however closed only to the Weibull distribution, brought about inappropriate application for the log-logistic observations. Proportional odds (PO) model in that case raised forward to perform similarly with the PH model. The subsequent modelling study is therefore producing a joint PO-longitudinal analysis rather than a widely applicable joint PH- longitudinal analysis. Latent parameters is introduced as a linkage technique between the two sub-models. Investigation in this study relies on the simulation statistics in which the survival time-to-event data and longitudinal measurements are both influenced by a covariate effects. The repeatedly measures data additionally allow for different kind of missingness mechanisms. Maximum likelihood estimation method is applied to the joint model parameters estimation. The performance of the joint model and separated sub-models are then be compared. The illustrated results contributed better estimators on the joint model instead of separated model.