Clayton Brown, PhD Candidate in Biostatistics
Generalized estimating equations (GEE) are formulated to approximate
group mean rate of change when there is informative drop-out (right
censoring.) The linear mixed model is employed to model continuous
response, and extension to the generalized mixed model for discrete
is also considered. Taylor series expansions are used to approximate the
first two moments to formulate quadratic estimating equations. Instead of
modelling the missing data mechanism, the fixed effects are
modelled as polynomial functions of drop-out time as a method to adjust for
informative missingness. This work extends and builds upon the "condi-
tional linear model" of Wu and Bailey (1989). The method is illustrated
with data from a clinical trial of a drug for treating Schizophrenia.
Simulations are also performed in the continuous case to compare various
estimators, including GEE.
Return to Longitudinal/Survival Working Group List | Return to Home Page