CFE 2019: Start Registration
View Submission - CMStatistics
Title: Sample size determination guidelines for latent mediation models: A Monte Carlo study Authors:  Junhao Pan - Sun Yat-sen University (China)
Jingheng Cai - Sun Yat-sen University (China)
Rongqian Sun - The Chinese University of Hong Kong (China) [presenting]
Abstract: Mediation analysis has been extensively applied to psychology and many other disciplines to explore the intrinsic influencing mechanism underlying phenomena or behaviors that researchers are interested in. An important issue researchers frequently encounter is: How many samples are needed to ensure sufficient power of testing the mediation effect?. Several empirical rules have been proposed for simple mediation models with manifest variables. Despite latent mediation analysis covers the measurement of unobservable constructs and is therefore much more widely used in practice, there is a paucity of study on the sample size required for testing latent mediation effects. The main objective is to provide empirically supported sample size guidelines for several representative latent mediation models commonly used in applied researches, which is helpful for experimental design and research funding application in psychology and so on. Monte Carlo simulation experiments are conducted to investigate minimum sample size required for power no less than 0.8 in the test of latent mediation effect under 144 modeling conditions constituted by influencing factors including the number of items per factor, effect size, and estimation methods. Besides, the comparison between frequentist and Bayesian approach is extended to the context of latent mediation analysis