On 17 Oct 2002, Sangdon Lee wrote: > I'm running a repeated measures of ANOVA using stat>>general linear > model>>repeated measures from SPSS v.10. I have one within-subject > effect (before and after workshop) and four between-subject effects > (sex, discipline, learning_style, and status). I have 24 subjects.
Twenty-four cases is not very many with four between-subject factors... > First run: When I specify the model with only all main and all two-way > interaction effects initially, I have significant effects for > within-subjects (workshop, workshop*status, > workshop*learning_style*gender) and for between-subjects > (learning_style, learning_style*gender, gender*discipline and > status*discipline). I'm confused... You stated that you ran only main effects and two-way interactions; yet you note a three-way interaction is significant... > Second run: Next I re-ran the model with the significant main and > two-way interaction effects identified in the first run. That is, the > within-subject effects is before and after workshop. The > between-subject effects are learning_style, status, gender, > discipline, learning_style*gender, gender*discipline, > status*discipline. What I found is that some main and interactions are > not significant anymore and also some main effects are significant > which was insignificant in the first run. I also found that depending > on which main and interactions are specified in the model, the > significant main and interaction effects changes. > > My question is why different results? Do I need to report the results > from the first run or second run? I appreciate any comments. Thanks. There is more than one error term in your design... If you omit effects (main effects and/or interactions) in the second run that were present in the first run, your error terms are going to change in ways that depend on the magnitude of the SS and df for the omitted effects... Without looking at the results, I can't really say more, but I am not at all surprised that the results differ... WBW __________________________________________________________________________ William B. Ware, Professor Educational Psychology, CB# 3500 Measurement, and Evaluation University of North Carolina PHONE (919)-962-7848 Chapel Hill, NC 27599-3500 FAX: (919)-962-1533 http://www.unc.edu/~wbware/ EMAIL: [EMAIL PROTECTED] __________________________________________________________________________ . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
