[Rasch] Insignificant MANOVA
Parisa Daftari Fard
pdaftaryfard at yahoo.com
Wed Jul 6 16:46:02 AEST 2016
Dear Mentors and Friends,
I have a question and I wonder if you can provide me with some published sources and references. In my research I ran MANOVA for four components of a questionnaire. The main effect turned out to be insignificant at p equals to 0.1; however, when I saw the between effect table (ANOVA) one of the variable was significantly different between groups. The editor of the journal following some statisticians like Pallant (2011) mentioned that I should not have reported the Anova table any more when the main effect in MANOVA is not significant. But he/she suggested me to run T-Test using Bonferroni adjustment (p = 0.0125). Considering the fact that the component turned out to be significant at p = 0.011 ((F (1, 102) = 6.736, p<0.05, r = 0.036), do you think that I should run T-test while the P meets the adjustment in ANOVA. When there is only one independent variable, there is no much difference between ANOVA and T-TEST. AM I right? Do you happen to have seen any article with insignificant M!
The Editor's Comment is as follows:
"You have used a MANOVA to compare the posttest scores and although the results of Multivariate tests are not significant you have referred to the table of Tests of Between-subjects effects and have marked some significant values. This is WRONG. When the results of Multivariate tests are not significant we are not allowed to go Tests of Between-subjects. See What Pallant (2011) says: "If you obtain a signiﬁcant result on this multivariate test of signiﬁcance, this gives you permission to investigate further in relation to each of your dependent variables" (p. 295) by checking tests of between subjects effects. If you insists on making comparisons on the posttest scores, you can run 4 separate t-tests by adjusting the Bonferroni adjustment by dividing .05 by 4 and gaining a new alpha level, that is .0125."
I should highly appreciate your kind attention and suggestion to solve this problem beforehand.
More information about the Rasch