I have one dependent, four independent and two moderator variables. Im testing moderation using hierarchical regression and need to test for multivariate normality.
Normality can be tested through different means {of course each one seperately}; of these is to test the skewness level, most of the text books accept skewness up to .5. you also can run scatter plot to check for normality. Box plot is another option, to exculed variable higher than 3 SD.
you meay also meant to check linearity, you can run scatter plot for each IV with the DV.
The Kolmogorov-Smirnov (n > 50) and Shapito-Wilks (n < 50) test give you information about if your variables are normally distributed. You can do this tests in SPSS by: Analyze >> Descriptive Statistics >> Explore >> button Graphics >> Select "Graphics with normality tests".
of course , it's better if you do some graphic to support these tests.
I think you would have to test the normality of errors after multiple regression, not the normality of all variables before.
Thank you prof. Ahmad. I did look at the kurtosis and skewness values and they are within range. But I understood them to be tests of univariate normality. Would they suffice as tests for multi variate normality as well?
One of the quickest ways to look at multivariate normality in SPSS is through a probability plot: either the quantile-quantile (Q-Q) plot, or the probability-probability (P-P) plot.
she asked for multivariate test and everybody answered her in terms of univariate normality test. her question was very valid and pertinent and i do not know why everybody answered like that. online statistics of web can do multivariate normality test for us. SPSS offers only Univariate test which is not correct to report if we have multivariate level data.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running