Main content

Marginal Approaches for Joint Models with Clustered Data

Show full item record

Title: Marginal Approaches for Joint Models with Clustered Data
Author: Gaudet, Danielle
Department: Department of Mathematics and Statistics
Program: Mathematics and Statistics
Advisor: Horrocks, JulieDarlington, Gerarda
Abstract: In the presence of clustering of individuals, standard errors (SEs) of model parameter estimates may be poorly estimated when models that assume independence of subjects are used. Marginal approaches in survival models aim to correct this, usually by estimating the model parameters under the assumption that all subjects are independent and then using a method that recognizes intra-cluster correlation to estimate the SEs. This thesis assesses the marginal approach with the application of group jackknife and group bootstrap to estimate SEs of parameter estimates in joint models with clustered data and compares them to model-based SEs that do not account for clustering. These methods were compared using a real data set. A simulation study compared the three methods of estimating the SEs against the empirical standard deviation (SD) of the parameter estimates. The results of the simulation study indicated that the group jackknife SEs and group bootstrap SEs were much closer to the empirical SDs than the model-based SEs from the misspecified joint model.
Date: 2021-12
Rights: Attribution-ShareAlike 4.0 International

Files in this item

Files Size Format View
Gaudet_Danielle_202112_MSc.pdf 730.6Kb PDF View/Open

This item appears in the following Collection(s)

Show full item record

Attribution-ShareAlike 4.0 International Except where otherwise noted, this item's license is described as Attribution-ShareAlike 4.0 International