topic Re: Multivariate normality in SAS Enterprise Guide

Multivariate normality

deleted_user — Thu, 19 Jul 2007 21:28:27 GMT

Hello!

I never used SAS.

I need to do Discriminant Analysis and MANOVA, and i can't understand what shall i do to get the multivariate normality test.

I've Learning Edition 2.0 (Enterprise Guide 2.1.40)... that i never used.

Can someone assist me on this with a very explicit manner, please?

Thanks in advance,
Américo Alves

Re: Multivariate normality

deleted_user — Fri, 20 Jul 2007 04:54:56 GMT

I doubt I can help you much.

First of all, I have EG3 and EG 4.1 installed on my computers, and so I can't direct you through the task lists for EG2. I presume they are available as options on the task lists and that is why you have named them, but I don't know to what extent they have changed in the past couple of years.

Secondly, deep and meaningful statistical discussions are something I leave to my namesake who would walk you through taking the code produced by EG and customising it to perform these analyses in a way that he would consider robustly reports the statistical analysis. That requires a detailed knowledge of both SAS and statistical procedures to perform the analysis and interpret the results. Note that the interpretation is not a "black box" and will involve formulation of hypotheses that you will test with a series of SAS procedures. It is something about which three days SAS training taught me a little, the greater part of what I learned is that it isn't something I would do alone.

I hope you have a statistician with you to assist in the interpretation of the results reported.

Kind regards

David

Re: Multivariate normality

deleted_user — Fri, 20 Jul 2007 13:05:32 GMT

Thanks for you comments, dkvj.

The interpretation of a multivariate normality test do not requires such a detailed knowledge. In a simplest way, the test produces one statistic and one p-value for it's significance. The null hypothesis is that the data set is similar to the normal distribution, therefore a sufficiently small p-value indicates non-normal data.

Henze-Zirkler test was recommended by my teachers as formal test of the multivariate normality, plus the use of Mardia's skewness and kurtosis measures.

Several friends told me that SAS can do it. I just don't know what shall i do using SAS to get this information.

If someone could assist me, it would be much appreciated.

Américo Alves

Re: Multivariate normality

ChrisHemedinger — Mon, 30 Jul 2007 17:19:12 GMT

Sounds like what you want is in the MODEL procedure. There is no point-and-click task to cover this, but you might be able to write your own PROC MODEL code to get what you need.

I'm not sure whether PROC MODEL is available in SAS Learning Edition 2.0 though. Not all of SAS' procedures are included in that version of Learning Edition -- you might have to use Learning Edition 4.1 or jump to the full version of SAS.

Chris