Thanks to all, for replying to my posting :
Let me answer all your questions :
I want to verify our theoretical asymptotic results with
the experiment for multinorm. So iam really looking for
a relaible test.
Yes Mardia 1970 (biometrica paper) and his book 1974 essentially
has the same asymptotic higher order moments results.
Iam presently using SAS macro, which claim to test multinorm
using Marida asymptotics with the data. Using it, as i posted in
my earlier message - i expect higher 'p-values' for both skewness
and kurtosis. Just because null hypothesis is to 'accept multi norm'.
It gives a high kurtosis (p value) and low skewness (p-value), Iam
not really clear as what going on.? Iam i missing something here.?
Iam not experinced with this multivariate testing. Any
advice/correction/pointer
would be of great help.
thanks for your time.
yoga
[EMAIL PROTECTED]
Glen Barnett wrote:
> yogab ?[EMAIL PROTECTED]? wrote in message
> [EMAIL PROTECTED]">news:[EMAIL PROTECTED]...
> ?
> ? Hello all.,
> ?
> ? Iam using mutivariate normality test (Mardia 1974) to check for
> ? normaility for my sample set of size 1000 (multivariate data).
> ?
> ? It seem to pass multivaiate kurtosis testing (higher p value)
> ? but consistently fails for skewness test. Iam presently using
> ? sas - 'multinorm' macro. I find this to be little hard to belive or
> ? rather
> ? if i can say little inconsistent.
> ?
> ? I just wanted to see if any one had the some experience with this test
> ? in particular or comments about mutivariate testing ? or any
> ? better way to do mutivariate normality testing ?
> ?
>
> I'm utterly confused. Why do you think it's inconsistent for it
> to reject on skewness and fail to reject on kurtosis?
>
> That sounds perfectly reasonable to me.
>
> Glen
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
