[Calendar Icon] Events Calendar

Dr. Giorgos Bakoyannis
Assistant Professor
Department of Biostatistics and Health Data Science
Indiana University Fairbanks School of Public Health

Nonparametric Analysis of Multistate Processes with Missing, Dependently Censored, and Clustered Data

Nonparametric analysis of multistate models with incomplete and clustered data is a challenging task and, to the best of our knowledge, has not been addressed so far. We propose a nonparametric inverse probability weighting estimator for state occupation probabilities of multistate processes with missing, dependently censored, and clustered data. The estimator does not impose Markov assumptions or assumptions regarding the structure of the within-cluster dependence and allows for informative cluster size. Closed-form variance estimators are provided and methodology for simultaneous confidence bands and nonparametric two-sample hypothesis testing is proposed. Using empirical process theory, we show that the estimator is uniformly consistent and converges weakly to a tight mean-zero Gaussian process. Moreover, we derive the asymptotic null distribution of the test statistic. Simulation studies show that the methods perform well even with a small number of clusters and a large percent of missingness. The proposed methods are used to analyze care transitions of HIV-infected individuals in a large multicenter study in East Africa.