A simulation-based approach to estimate joint model of longitudinal and event–time data with many missing longitudinal observations

Joint models of longitudinal and event–time data have been extensively studied and applied to many fields. Estimation of joint models is challenging, existing procedures are computational expensive and their effectiveness relies heavily on the data quality. In this study, a novel simulation-based procedure is proposed to estimate a general family of joint models, which include many widely applied joint models as special cases. Our procedure can easily handle low-quality data where longitudinal observations are systematically missing for some of the covariate dimensions. In addition, our estimation procedure is implementable through parallel computing, so it is perfectly applicable to massive data in finance. The consistency and asymptotic normality of our estimator are proved, numerical experiments with both synthetic datasets and a real consumer–loan dataset are carried out to illustrate the effectiveness of the procedure.