Dear All,
I wish to find a linear regression model using my continuous response
(y). Other independent variables are binary.
My response is in the form of percentages and unfortunately histograms
and box-plots have shown its distribution to be higly skewed
(non-normal). The transformations I have tried have not produced very
successful changes in the distribution.
I believe there's a transformation available that work's well on
percentages.
Does anyone have any ideas or references as to what it might be?
Thanks
Susan White
|