1. Introduction.- I: Logistic Regression with Two Categorical Covariates.- 2. The complete data case.- 3. Missing value mechanisms.- 4.
Estimation methods.- 4.1 Maximum Likelihood (ML) Estimatio.- 4.2 Pseudo Maximum Likelihood (PML) Estimatio.- 4.3 The Filling metho.- 4.
4 Complete Case Analysi.- 4.5 Additional Categor.- 4.6 Probability Imputatio.- 4.7 Omission of Covariat.- 5.
Quantitative comparisons: Asymptotic results.- 5.1 Asymptotic relative efficiency: ML Estimation vs. PML Estimatio.- 5.2 Asymptotic relative efficiency: ML Estimation vs. Fillin.- 5.
3 Asymptotic relative efficiency: PML Estimation vs. Fillin.- 5.4 Asymptotic relative efficiency: ML Estimation vs. Complete Case Analysi.- 5.5 Asymptotic relative efficiency: ML Estimation for complete data vs. ML Estimation for incomplete dat.
- 5.6 Asymptotic relative efficiency: ML Estimation for complete data vs. Complete Case Analysi.- 5.7 Asymptotic relative efficiency: A summary of result.- 5.8 Asymptotic bias: Comparison of Probability Imputation, Additional Category and Omission of Covariat.- 5.
9 Asymptotic bias: Evaluation of Conditional Probability Imputatio.- 5.10 Evaluating the underestimation of variance of Conditional Probability Imputatio.- 5.11 The importance of the variance correction of the Filling metho.- 6. Quantitative comparisons: Results from finite sample size simulation studies.- 6.
1 Finite behavior of ML Estimation, PML Estimation, Filling and Complete Case Analysi.- 6.2 Power comparison.- 6.3 Evaluation of Conditional Probability Imputatio.- 7. Examples.- 7.
1 Illustrating artificial example.- 7.2 An example with a real data se.- 8. Sensitivity analysis.- II: Generalizations.- 9. General regression models with missing values in one of two covariates.
- 9.1 ML Estimatio.- 9.2 Semiparametric ML Estimatio.- 9.3 Estimation of the Score Functio.- 9.4 Complete Case Analysi.
- 9.5 Mean Imputation and Additional Categor.- 9.6 The Cox proportional hazards mode.- 10. Generalizations for more than two covariates.- 10.1 One covariate with missing value.
- 10.2 Missing values in more than one covariat.- 11. Missing values and subsampling.- 11.1 Two stage design.- 11.2 Surrogate covariates and validation samplin.
- 11.3 Subsampling of the nonresponder.- 11.4 (Sub-)sampling of additional variable.- 12. Further Examples.- 12.1 Example 1: Risk factors for subsequent contralateral breast cance.
- 12.2 Example 2: A study on the role of DNA content for the prognosis of ovarian cancer patient.- 13. Discussion.- 13.1 Statistical inference if the MAR assumption is satisfie.- 13.2 Statistical inference if the MAR assumption is questionabl.
- 13.3 Topics of future researc.- 13.4 Final remar.- Appendices.- A. 1 ML Estimation in the presence of missing values A.2 The EM algorithm.
- B. 1 Explicit representation of the score function of ML Estimation and the information matrix in the complete data case.- B. 2 Explicit representation of the score function of ML Estimation and the information matrix.- B. 3 Explicit representation of the quantities used for the asymptotic variance of the PML estimates.- B. 4 Explicit representation of the quantities used for the asymptotic variance of the estimates of the Filling method.
- References.- Notation Index.