Research Seminar Marketing - "Mind Your Zee: More Insights in the Effect Size, Observed Power, and Required Sample Size for Replicable Mediation Analyses from z-values" - Rik Pieters, Tilburg University

Statistical mediation analysis is a leading tool in marketing, organizational behavior, and social psychology to gain insights in the process by which a treatment or characteristic of people or organizations influences downstream consequences via mediating variables. The results of these analyses are predominantly and almost exclusively reported as the resampled 95% confidence interval of the mediated effect. The mediation is then considered present-or-absent when this interval does not or does cover the zero. This project aims to show that, although confidence intervals are fine, the practice of relying almost solely on them to make present-absent decisions is counterproductive, that it increases the likelihood of identifying mediation effects that are not replicable, prevents deeper insights into the mediation process, its effect size, observed power, and the required sample size for the planned next study. It proposes the z-value of the mediated effect, obtained from Monte Carlo analyses, as a simple metric to account for these issues. It reanalyzes classic mediation studies, reports on a simulation study, and on a small-scale re-analysis of recent mediation analyses in the Journal of Consumer Research. It offers a R-library and accessible shinyApp to conduct the analyses using sufficient statistic data only. Time permitting, the presentation may explore the possibilities of a generalized theory of apple pie home-production and consumption. Join Zoom Meeting https://unil.zoom.us/j/8466956805?omn=96911854880 Meeting ID: 846 695 6805 Password: happy2024