Though I am not a statistician, I think I know this.
The dependent variable needn't be unimodal or normal. What should be
unimodal and normal are the residuals (the errors) remaining after fitting
the multiple linear regression model. If they are not, consider either
another model, or transform the dependent variable.
John Thaden, Ph.D.
Biochemist
University of Arkansas for Medical Sciences
Little Rock, Arkansas USA
At 12:11 AM 5/22/1998 +0800, Dr. Hin wrote:
>Please forgive this question of mine that is more statistical than S
>orientated...
>
>We know that if the dependent variable in multiple regression should be a
>unimodal, normal distribution. In the event that the dependent variable is
>a bimodal, or multimodal distribution, and a multivariate analysis is
>necessary, what strategies would you suggest? Can I still use linear
>regression, or do I need to transform the variable (knowingly, and accept
>that it will lose information)?
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
|