I currently have one IV with two levels (married or divorced) and my DV is based upon a Likert scale of dimensions, therefore I have 5 DV's. One of my committee members recommended to use independent t-tests, while the other stated I should use MANOVA.
The MANOVA is a concern because (per Laerd) it "combines the two or more dependent variables to form a 'new' dependent variable in such a way as to maximize the differences between the groups of the independent variable." So would this result in all 5 of my DV's being combined? I am confused.
Also, because I am using individual statistical tests do I reduce my alpha level to account for experimentwise error (0.05/5 tests = 0.01) when using independent t-tests and or MANOVA?
Thanks for your help!
Dear Angelique,
please can you say what means IV and DV?
Best regards
Tanja
- Hi,
- Very simply, if the DVs are related/correlated you can use MANOVA, if they are independent then just use t-test.
Dear Angelique,
the requrement for the MANOVA is, that the parameters are normally distributed. Because you want to compare Likert scales, the parameters are not normally distributed. I think, the member of the committee know this, too - but there is no other test. Therefore, I have not a reliable answer for your problem.
Best regard
Tanja
Dear Angelique,
I think either method would be acceptable and your Bonferroni correction with the significance level for the t test is sensible.
I agree with Tanja that you need to check the assumptions of MANOVA but normality is not necessarily required because of robust exceptions. Please refer to a reputable text book, such as Field. I'd suggest trying this first and keeping multiple t tests as your backup strategy.
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude