|
Contents 1 Introduction 1 1.1 Clustered Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Types of Correlated Data . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.3 Statistical Models for Correlated Data . . . . . . . . . . . . . . . . . . 3 1.4 Organization of Subsequent Chapters . . . . . . . . . . . . . . . . . . . 6 2 Introduction to Multilevel Models 11 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.2 The 1997 Belgian Health Interview Survey . . . . . . . . . . . . . . . . 12 2.3 Linear Multilevel Models . . . . . . . . . . . . . . . . . . . . . . . . . . 13 2.4 Nonlinear Multilevel Models . . . . . . . . . . . . . . . . . . . . . . . . 16 2.4.1 Maximum Marginal Likelihood . . . . . . . . . . . . . . . . . . 16 2.4.2 Approximate Methods . . . . . . . . . . . . . . . . . . . . . . . 18 2.5 Weighting in Multilevel Models . . . . . . . . . . . . . . . . . . . . . . 19 2.6 Application to the HIS . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.6.1 Linear Multilevel Model . . . . . . . . . . . . . . . . . . . . . . 21 2.6.2 Multilevel Logistic Model . . . . . . . . . . . . . . . . . . . . . 22 3 Pairwise Likelihood Estimation in Multilevel Probit Models 27 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.2 Pseudo-Likelihood Estimation . . . . . . . . . . . . . . . . . . . . . . . 28 3.2.1 Pseudo-Likelihood De nition . . . . . . . . . . . . . . . . . . . 29 3.2.2 Asymptotic Properties of Pseudo-Likelihood Estimators . . . . 30 3.3 Pairwise Likelihood in the Multilevel Probit Model . . . . . . . . . . 37 3.3.1 The Multilevel Probit Model . . . . . . . . . . . . . . . . . . . 38 3.3.2 Pairwise Likelihood . . . . . . . . . . . . . . . . . . . . . . . . 39 3.4 Asymptotic Relative Eciency . . . . . . . . . . . . . . . . . . . . . . 41 3.5 Weighted Pairwise Likelihood . . . . . . . . . . . . . . . . . . . . . . . 45 3.6 Example: a Meta-Analysis of Trials in Schizophrenic Subjects . . . . . 49 3.7 Simulation Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 3.8 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 4 Validation of Surrogate Endpoints in Multiple Randomized Clinical Trials with Discrete Outcomes 65 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 4.2 Surrogate Endpoint Validation: Two Normally Distributed Endpoints 67 4.2.1 A Hierarchical Model . . . . . . . . . . . . . . . . . . . . . . . 67 4.2.2 Trial-Level Surrogacy . . . . . . . . . . . . . . . . . . . . . . . 68 4.2.3 Individual-Level Surrogacy . . . . . . . . . . . . . . . . . . . . 70 4.2.4 Surrogate Evaluation . . . . . . . . . . . . . . . . . . . . . . . . 71 4.2.5 Computational Issues . . . . . . . . . . . . . . . . . . . . . . . 71 4.3 Surrogate Endpoint Validation: Two Binary Outcomes . . . . . . . . 73 4.3.1 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 4.3.2 Model Estimation . . . . . . . . . . . . . . . . . . . . . . . . . 73 4.4 Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 4.5 Example: a Meta-Analysis of Trials in Schizophrenic Subjects . . . . . 78 4.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 5 Repeated-Measures Models to Evaluate a Hepatitis B Vaccination Program 85 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 5.2 Hepatitis B Vaccination Program and Scienti c Questions . . . . . . . 86 5.3 The Linear Mixed Model With Serial Correlation . . . . . . . . . . . . 89 5.4 Fractional Polynomials with Longitudinal Data . . . . . . . . . . . . . 91 5.5 Time-evolution of Antibodies . . . . . . . . . . . . . . . . . . . . . . . 93 5.6 Prediction at Year 12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 5.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 6 Estimating Reliability Using Non-Linear Mixed Models With Re- peated Binary Data 105 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 6.2 Estimating Reliability in Generalized Linear Mixed Models . . . . . . 107 6.2.1 General Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 6.2.2 Probit Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 6.3 Estimating Reliability in the Probit Model with Autocorrelation . . . 110 6.3.1 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 6.3.2 Model Estimation . . . . . . . . . . . . . . . . . . . . . . . . . 112 6.3.3 Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 6.3.4 Application to the Schizophrenia Data . . . . . . . . . . . . . . 116 6.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 7 Validation of a Longitudinally Measured Surrogate Marker for a Time-to-Event Endpoint 119 7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 7.2 Motivating Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 7.3 Modeling Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 7.4 Application to the Advanced Prostate Cancer Data . . . . . . . . . . . 127 7.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 8 Concluding Remarks and Further Research 133 8.1 Pairwise Likelihood Estimation . . . . . . . . . . . . . . . . . . . . . . 133 8.1.1 Model Checking and Diagnostics . . . . . . . . . . . . . . . . . 134 8.1.2 Missing Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 8.1.3 Crossed Random-Eects Models . . . . . . . . . . . . . . . . . 136 8.2 Evaluation of Surrogate Endpoints . . . . . . . . . . . . . . . . . . . . 138
|