Multilevel structural equation modeling for 2-1-1 mediation using Bayesian method
No Thumbnail Available
Authors
Meeting name
Sponsors
Date
Journal Title
Format
Thesis
Subject
Abstract
The multilevel mediation effect from the 2-1-1 designed model was analyzed using the Bayesian method. For Bayesian analysis, eight default priors and one informative prior were used. The priors for the location and the variance parameters associated with the mediation effect were diversely specified and compared in terms of how the estimated mediation effect changed depending on the prior distribution. Findings are analyzed in terms of convergence, relative bias, RMSE, coverage rates, and power for all location and variance parameters. Furthermore, based on different prior distributions, the data conditions that best estimated the multilevel mediation effect were identified. Followed by the simulation findings, a sensitivity analysis was conducted using this empirical example to determine the robustness of the results through various analyses. The prior sensitivity analysis reveals that the estimate of the indirect parameter was robust when the EB priors that contained even a small amount of data information or the informative prior were selected. The findings from this study have implications for specification of EB priors, selection of preferred methods, and appropriate data conditions when using multilevel mediation model. Further investigation of extended models, diverse data conditions, and different types of prior conditions is warranted.
Table of Contents
PubMed ID
Degree
Ph. D.