Analysis of Incomplete Data from Longitudinal and Mixed Longitudinal Studies

SUMMARY An attempt has been made in this paper to outline methods of analysis which may be applied to incomplete longitudinal data. A specific example dealing with the estimation of a mean vector has served as the vehicle of comparison of two methods of analysis: (a) a mixed longitudinal analysis due originally to Rao and Rao (1966) and (b) a cross-sectional type of analysis. The mixed longitudinal model has been framed within the context of a more general model proposed by Kleinbaum (1973). By simplifying the model in a special case to the usual univariate linear model, the methods described may be carried out on conventional linear model computer programs. Limitations of the method of analysis are discussed.