https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8967310/
MacKinnon, D. P., & Lamp, S. J. (2021). A unification of mediator, confounder, and collider effects. Prevention Science, 22(8), 1185–1193. https://doi.org/10.1007/s11121-021-01268-xMacKinnon, D. P., & Lamp, S. J. (2021). A unification of mediator, confounder, and collider effects. Prevention Science, 22(8), 1185–1193.
https://doi.org/10.1007/s11121-021-01268-x
QT:{{”
Third-variable effects are not distinguishable solely by statistical methods. Each third-variable effect can be fit to the same data, and if the relations between the variables are substantial, there will be evidence for each effect. In this sense, the confounder, mediator, and collider models are equivalent, providing an equal representation of the information contained in the data for three variables (Stelzl, 1986). Although mediation, confounding, and collision may equally explain the statistical associations among three variables, they describe different causal relations among those variables. Like much recent research on causal analysis, this paper highlights the centrality of the causal model underlying a research study and the important distinction between the causal model and the statistical model. The appropriate causal model is determined by prior empirical research and theory. The statistical analysis provides estimates for the proposed causal model.
“}}
linkstream2 microblog 