Assumption of Manova?

#1
I have a researh 4 IV where 3 of it are categorical type and 1 is continuous, to 3 DV.
Is anyone here know about normality assumption of Manova? What we should use in normality test :
1. Normality only for dependent variable? or,
2. Normality in relation between dependent variable with the categorical of IV?

Thanks
 

obh

Active Member
#2
Generally in ANOVA or Regression the normality assumption is for the residuals.
So I assume the same for MANOVA. (the same as linear regression)

ANOVA test shouldn't be sensitive for moderate deviation from the normality when the sample size is big enough (CLT)
 
#3
Generally in ANOVA or Regression the normality assumption is for the residuals.
So I assume the same for MANOVA. (the same as linear regression)

ANOVA test shouldn't be sensitive for moderate deviation from the normality when the sample size is big enough (CLT)
Could you give more detail explanation?
I have 3 categories of IV, it means there are 3 crosstabs categories with 30 sampels in each.
So, should I do the normality test?
 

obh

Active Member
#4
Hi Ghoza,

In MANOVA the assumption is for multivariate normality, which is difficult to check.
You may check separately the normality of each DV for the IVs, not as good bu practically used.

but if you have at list 20 observations for each combination of dependent*independent variable. you don't need to check for multivariate normality. (or 20 DF in the univariate F test)
 
#5
Hi Ghoza,

In MANOVA the assumption is for multivariate normality, but if you have at list 20 observations for each combination of dependent*independent variable. you don't need to check for multivariate normality. (or 20 DF in the univariate F test)
Oh okay I understand. Do you have the references of this? Maybe in a Book or statistical articles?
Thanks for the answer btw.