University of Illinois Urbana-Champaign

A Bayesian approach to crossed-random-effects mediation analysis for zero-inflated mediators and binary outcomes

Pan, Pei-Yu (Marian)

Permalink

https://hdl.handle.net/2142/110639

Description

Title
A Bayesian approach to crossed-random-effects mediation analysis for zero-inflated mediators and binary outcomes
Author(s)
Pan, Pei-Yu (Marian)
Issue Date
2021-03-24
Director of Research (if dissertation) or Advisor (if thesis)
Anderson, Carolyn Jane
Committee Member(s)
  • Anderson, Richard C
  • Jiang, Gabriella
Department of Study
Educational Psychology
Discipline
Educational Psychology
Degree Granting Institution
University of Illinois at Urbana-Champaign
Degree Name
M.S.
Degree Level
Thesis
Keyword(s)
Bayesian estimation, mediation, crossed random effects, zero-inflated mediator, academic vocabulary
Abstract
In crossed random effects designs, observations are nested in the combination of two random factors, e.g., subjects and stimuli. Such designs are popular in experimental research in social sciences. Crossed random effects models (CREM) can accommodate the random effects of both subjects and stimuli. Based on academic vocabulary research, the model of interest in the current study was a mediation model with crossed random effects, a zero-inflated mediator, and a binary outcome. With maximum likelihood estimation (MLE), the mediation model could not converge, which was consistent with the previous finding on analyzing CREM with MLE (Huang & Anderson, 2020). Therefore, the current study investigated whether Bayesian estimation can be a viable alternative as suggested by previous research. The simulation results indicated that Bayesian estimates were essentially unbiased and precise. There were only two out of 180 models did not converge; future studies can increase the iterations or use an informative prior to resolving the non-convergence issue. Moreover, a Bayesian method was developed to compute the mediation effect; the results of the Bayesian method closely aligned with the bootstrapping results. Lastly, the application of Bayesian estimation was demonstrated in an academic vocabulary study.
Graduation Semester
2021-05
Type of Resource
Thesis
Permalink
http://hdl.handle.net/2142/110639
Copyright and License Information
(Copyright 2021 Pei-Yu (Marian) Pan)

Owning Collections

Graduate Dissertations and Theses at Illinois PRIMARY
Graduate Theses and Dissertations at Illinois
Dissertations and Theses - Education
Dissertations and Theses from the College of Education