Dr. Wei Liu

Department of Mathematics and Statistics

York University

Title: "Semiparametric Nonlinear Mixed-effects Models with Covariate Measurement Errors and Missing Responses"

HIV viral dynamic models have received great attention in recent years. These models approximate
the viral load trajectories after the start of an anti-HIV treatment. The viral dynamic parameters
can be used to evaluate the efficacy of the anti-HIV treatment. Parametric nonlinear mixed-effects
(NLME) models have been widely used to model HIV viral dynamics in the initial period. After the
initial period, however, parametric models may not fit the data well and a semiparametric or
nonparametric model may be more flexible. Therefore, a semiparametric NLME model may be useful for
modeling long-term HIV viral dynamics. In practice, statistical analyses of HIV viral dynamics are
complicated by the following problems: (i) important covariates such as CD4 cell count are often
measured with substantial errors; (ii) missing data often occur in the response (viral load)
measurements due to patients' dropouts or other problems, and the missing data are likely to be
non-ignorable in the sense that the missingness may be related to the missing values; (iii) missing
data may also occur in time-varying covariates due to different measurement schedules or other
problems. We consider likelihood methods which simultaneously address measurement error and missing
data problems in semiparametric NLME models. A real HIV dataset is analyzed in detail, and simulation
studies are conducted to evaluate the methods.