LISTSERV at the University of Georgia
Menubar Imagemap
Home Browse Manage Request Manuals Register
Previous messageNext messagePrevious in topicNext in topicPrevious by same authorNext by same authorPrevious page (September 2005)Back to main SPSSX-L pageJoin or leave SPSSX-L (or change settings)ReplyPost a new messageSearchProportional fontNon-proportional font
Date:         Fri, 2 Sep 2005 20:52:00 +0100
Reply-To:     starborn@statisticsdoc.com
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Keith Starborn <starborn@statisticsdoc.com>
Subject:      Re: MANOVA vs. Repeated Measures ANOVA
Comments: To: "MacLennan, Graeme" <g.maclennan@ABDN.AC.UK>
Content-Type: text/plain; charset=ISO-8859-1

Keith Starborn www.statisticsdoc.com

Dear Graeme,

In the situation that you describe, the ANCOVA can be viewed as a special case of the ANOVA in which the pre-treatment measure covariate is entered as a covariate before the treatment group variable, and the increment in prediction of the outcome is assessed after adjusting for the covariate. There are certainly cases where ANCOVA would not be appropriate. ANCOVA assumes that the relationship between the covariate and the outcome is equal in the two groups (i.e., there is no group by covariate interaction, and thus there is homgeneity in the regression effect of the covariate on the outcome). If the relationship between the covariate and outcome is not equal across groups, then the GROUP by COVARIATE effect on the outcome interaction must be entered into the analysis, and the unique effect of the treatment group factor would be considered after adjusting for both the covariate and the interaction.

If ANCOVA is used more widely in medical fields, I would like to think that this happens because the assumption of homogeneity of covariate effects is supported by a substantive body of research. I have seen a lot of ANCOVA in psychological and educational research, especially program evaluations, where I am not so sure that this assumption is supportable.

Best,

KS

Quoting "MacLennan, Graeme" <g.maclennan@ABDN.AC.UK>:

> Dear listers, as an aside to this thread, I would always choose ANCOVA over > repeated measures ANOVA for pre to post intervention data(that is only two > time points). As a medical statistician ANCOVA is a fairly bog standard > tool, but as I work more with psychologists I see repeated measures ANOVA > used in its stead for pre test post test scores set up, analogous to a > randomised controlled trial in medicine. Is there a real divide on this > issue or am I imagining it? We medical statos use ANCOVA because we are told > to, see Vickers AJ, Altman DG. Bmj 2001: 323(7321); 1123-4, Senn S. > Statistics in Medicine 2000: 19; 861-877, but I would like to see any > arguments in the literature to the contrary. > > I look forward to any views on this. > > Kind regards, Graeme. > > -----Original Message----- > From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of > Keith Starborn > Sent: 01 September 2005 20:43 > To: SPSSX-L@LISTSERV.UGA.EDU > Subject: Re: MANOVA vs. Repeated Measures ANOVA > > Keith Starborn > www.statisticsdoc.com > > Dear Lana, > > Dr. Swanks' suggestion regarding Huberty and Morris (1989) is an excellent > one > (as is the reference to Tabachnick & Fidellas by Dr. Buchanan). Many > journal > editors, and an entire generation of graduate students in Education and > behavioral sciences, have been socialized to believe that a significant > MANOVA > is a necessary condition for conducting ANOVAs, that MANOVAs are needed > whenever a study has more than one dependent variable in order to guard > against > inflated alpha levels, and that the need for MANOVA is even greater when the > dependent variables are correlated. These assumptions have been disputed > most > notably by Huberty and Morris (1989) and others. > > Having said that, you may or may not glean some interesting insights > from asking > multivariate questions of your data, along the lines discussed by Huberty > and > Morris. Do you consider your set of five dependent variables to measure a > complex of related school adjustment variables? Are the program activities > ones that you would expect to impact on this set of school adjustment > variables? Are you perhaps interested in the multivariate structure of the > relationships between program actitivities and change in multiple school > adjustment indices? > > HTH, > > KS > > Quoting "Yampolskaya, Svetlana" <yampol@fmhi.usf.edu>: > >> Dear List, >> >> I have 5 DV (i.e., student GPA, number of disciplinary referrals, days a >> student was absent from school, etc.). These DVs were measures twice >> (the beginning of the semester and at the end). My IV is level of >> participation in program activiteis. I used Repeated Measures ANOVA. >> The reviewers of the manuscript made the following comment: >> >> Why MANOVAs (vs ANOVA) were not conducted to account for the >> correlations (relatedness) between dependent variables? >> >> Can anybody help and explain when it is more appropriate to use MANOVA >> vs ANOVA? >> >> Thank you in advance, >> >> Lana >> > > > > -- > For personalized and experienced statistical consulting, > visit www.statisticsdoc.com >

-- For personalized and experienced statistical consulting, visit www.statisticsdoc.com


Back to: Top of message | Previous page | Main SPSSX-L page