Precision of testing the indirect effect for mediation studies
thesis
posted on 2011-07-18, 00:00authored byMelissa Ann Mitchell
Sample size planning is often one of the first steps when planning a psychological study. Researchers generally need to know how many participants they will need to have 0.80 power. Recently, methodologists have taken another view on sample size planning, mainly planning a study to have an accurate and precise estimate (AIPE). This work focuses on the latter view on sample size planning for mediation studies (how large must a sample be to have an accurate estimate of the indirect effect). This work looks at four methods to assess mediation in two different models: a cross-sectional and longitudinal model. Single sample tests require the more participants than the resampling approaches in both models. Coverage rates vary between methods in the models but it appears that the percentile bootstrap has the best coverage rates. We also provide R and MPlus codes for researchers to examine parameter values we did not examine in this study so they can plan studies with adequate precision.