I always thought that one benefit of using MANOVA is to avoid inflating the risk of Type 1 error that might occur by conducting, say, multiple t-tests. However, I’ve since found a few places where people advise using a correction such as Bonferroni to alpha values of the univariate tests that follow the multivariate test. Can anyone point me in the right direction to get a [semi] definitive answer? (I’ve got 4 DVs.)
Dear Andrew,
It depends on your model. Bonferoni corrections are also available in other statistical tests like anova. However normally you follow-up a MANOVA with discriminant analysis. However your biggest problem with MANOVA is the assumptions. Take great care in checking them.
Good luck
Suppose you have to compare the values of 2 variables (let's say for instance systolic and diastolic blood pressure) between 4 groups of patients (A, B, C and D).
MANOVA avoids the multiplicity correction issues for comparing both SBP and DBP, by using correlations between them and giving a single p-value. If significant, it will mean « there is at least one of the 4 groups in which either SBP, DBP or both is different from the others ».
For going further, you'll still have to do univariate or multivariate T-tests between the groups and variables of interest. For this step, multiplicity issues will arise; hence, Bonferroni correction could be used. Note that better method also exists, especially if MANOVA assumptions are met.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA