Simpler Approach for Mediation Analysis for Dichotomous Mediators in Logistic Regression
Document Type
Presentation
Presentation Date
3-9-2016
Abstract or Description
Mediation is a hypothesized causal chain in which one variable affects a second that, in turn, affects a third. It mediates the relationship between predictors and outcomes. To select and test for a potential mediator, the potential mediator should be associated with the predictor variable and with the outcome variable and should lie in the causal pathway between the predictor and the response. The mediation analysis for continuous response variables is well developed in the literature and it can be shown that the total effect of(X)on(Y),c, is equal to c’+ab, where ab is the mediation effect of the variable(M). However, for categorical responses mediation analysis still not fully developed. In this paper, we propose and developed a new approach using the latent variable technique to adjust for c=c’+ab. Our intensive simulation study and theoretical developments showed that on average the proportion of the mediation effect of using our latent variable approach relative to direct approach is about 0.412. Real data example is used to illustrate the proposed approach.
Sponsorship/Conference/Institution
Eastern North American Region International Biometric Society Spring Meeting (ENAR)
Location
Austin, TX
Recommended Citation
Samawi, Hani M., Jingxian Cai, Haresh Rochani, Daniel F. Linder.
2016.
"Simpler Approach for Mediation Analysis for Dichotomous Mediators in Logistic Regression."
Biostatistics Faculty Presentations.
Presentation 2.
https://digitalcommons.georgiasouthern.edu/biostat-facpres/2
