| In this thesis, we study the problem of making inference on the treatment effect in a smoking cessation trial where the design calls for the measurement of the primary outcome and its surrogate, and the outcomes are subject to missing values.;A dynamic generalized linear mixed effects model linking the outcomes and the covariates is proposed. Assuming logistic regression for the primary outcome, either normal or baseline logit model for the surrogate outcome, the maximum likelihood approach is first applied to the estimation and inference on the model parameters under the missing at random assumption. The average treatment effect is then estimated by the G-computation approach. The proposed approach can handle missing outcomes that form arbitrary missing patterns over time. The performance of the proposed method is evaluated by a simulation study that mimics the smoking cessation data.;We show that the proposed model is superior to the model ignoring surrogate outcome, so the surrogate outcome is valuable in recovering information on the primary outcome. |
