Re: Statistical question-determining mediation

[Stanley Cohen]

See the Sobel test:   http://www.unc.edu/~preacher/sobel/sobel.htm   Stan   *~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~* "Every day, computers are making people easier to use."                               - David Tempkin                                 Motto from In Formation    Stanley H. Cohen, PhD Professor Past Faculty Senate Chair, 2002-2003 Department of Psychology West Virginia University 1220 Life Sciences Building Morgantown, WV 26506-6040 E-Mail: [EMAIL PROTECTED] http://www.as.wvu.edu/psyc/Faculty/SCohen/index.htm Tel. No.: 304-293-2001x31641 Fax: 304-293-6606 *~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~* >>> [EMAIL PROTECTED] 2/17/2006 1:23 PM >>> I have a statistical question about mediation. It is not exactly teaching-related although I am trying to teach my students how to do it. I hope one of you can help. So I'm following Baron and Kenny's (1986, JPSP) instructions for how to to test a mediational model. It is quite simple. The IV is prior dating violence experience ("experience"), the DV is the personal risk estimate for experiencing dating violence in the next 10 years ("risk"), and the mediator is depression. So according to Baron and Kenny you must first regress the mediator onto the IV. That is indeed significant. Second step is to regress the DV onto the IV. That is also significant. The standardized beta is .218, t = 2.88, p = .005. In the third step I regress the DV onto both the IV and the mediator. The standardizes beta for the IV (experience) is .167, t = 2.19, p =.03. So clearly the beta is lower but still significant. So now my question. Did depression mediate the relationship between experience and risk? I am not interested in using a SEM (so no need to suggest that). Baron and Kenny say there is no need to do a partial correlation or a hierarchical regression model but they do not say exactly how you determine mediation. I understand Cohen and Cohen have a test for comparing the two beta values but that the test is quite a powerhog (I only have 193 Ss). Cheers Marie -- ********************************************* Marie Helweg-Larsen, Ph.D. Associate Professor of Psychology Dickinson College, P.O. Box 1773 Carlisle, PA 17013 Office: (717) 245-1562, Fax: (717) 245-1971 Webpage: www.dickinson.edu/~helwegm ********************************************* --- You are currently subscribed to tips as: [EMAIL PROTECTED] To unsubscribe send a blank email to [EMAIL PROTECTED] --- You are currently subscribed to tips as: archive@jab.org To unsubscribe send a blank email to [EMAIL PROTECTED]