Date on Master's Thesis/Doctoral Dissertation
5-2016
Document Type
Master's Thesis
Degree Name
M.S.
Department
Bioinformatics and Biostatistics
Degree Program
Biostatistics, MS
Committee Chair
Lorenz, Douglas
Committee Member
Datta, Somnath
Committee Member
Terson de Paleville, Daniela
Abstract
The log rank test is a popular nonparametric test for comparing the marginal survival distribution of two groups. When data are organized within clusters and the size of clusters or the distribution of group membership within a cluster is related to an outcome of interest, traditional methods of data analysis can be biased. In this thesis, we develop a within-cluster group weighted log rank test to compare marginal survival time distributions between groups from clustered data, correcting for cluster size and intra-cluster group size informativeness. The performance of this new test is compared with the unweighted and cluster-weighted log rank tests via a simulation study. The simulation results suggest the new test performs appropriately under scenarios of cluster size and intra-cluster group size informativeness, and produces higher power than the two comparison tests under non-informative scenarios. The new test is then illustrated on a live data set comparing time to functional improvement in task performance from patients with spinal cord injuries.
Recommended Citation
Gregg, Mary Elizabeth, "A log rank test for clustered data under informative within-cluster group size." (2016). Electronic Theses and Dissertations. Paper 2434.
https://doi.org/10.18297/etd/2434
