Joint Modeling of Measurements and Missingness

The problem of dealing with missing values is common throughout statistical work and is almost ever present in the analysis of longitudinal or repeated measurements data. Missing data are indeed common in clinical trials (Piantadosi 1997, Green, Benedetti, and Crowley 1997, Friedman, Furberg, and DeMets 1998), in epidemiologic studies (Kahn and Sempos 1989, Clayton and Hills 1993, Lilienfeld and Stolley 1994, Selvin 1996), and, very prominently, in sample surveys (Fowler 1988, Schafer, Khare and Ezatti-Rice 1993, Rubin 1987, Rubin, Stern, and Vehovar 1995).

Patients who drop out of a clinical trial are usually listed on a separate withdrawal sheet of the case record form with the reasons for withdrawal, entered by the investigator. Reasons typically encountered are adverse events, illness not related to study medication, uncooperative patient, protocol violation, ineffective study medication, and other reasons (with further specification, e.g., lost to follow-up). Based on such a medical typology, Gould (1980) proposed specific methods to handle this type of incompleteness.