While these packages are all very useful, there was still a need for a package enabling mediation analysis within the framework of structural equation modelling (SEM).<br>More specifically, we needed a package which could estimate all the regression models simultaneously (as opposed to the traditional series of sequential regression analyses) including observed or/and latent variables (as opposed to models including only observed variables). In this paper, I would like to present the medsem command (developed using Stata’s programming language) which does exactly this task. This is a post estimation command typed in after the estimation of a mediation model using the Stata’s built-in sem command for SEM.<br>The paper is structured in the following way. First, in the subsequent Section 2, mediation analysis is defined and explained based on Baron and Kenny’s (1986) approach including the two procedures based upon which the medsem command is programmed. In Section 3, applications of two relatively simple and a rather complex mediational model are presented to demonstrate the performance of the medsem command. Finally, in the conclusion section, the benefits, limitations, as well as future development possibilities associated with the command, are mentioned.
正在翻譯中..
