1

I am trying to run a simple regression model between a couple of variables, one of which is bound between values of (.25 - .75). This will be my target variable. I understand the beta distribution can take values between (0 and 1), but I am not sure I should use beta regression modeling. Does anyone have any advice on the modeling approach?

James
  • 21
  • 2
  • 1
    There is a paper called [A better lemon squeezer](http://psychology3.anu.edu.au/people/smithson/details/Pubs/Smithson_Verkuilen06.pdf) that may be of use. It covers the uses of beta regression in several situations and, if I recall, some of them are like yours. – Peter Flom Oct 22 '14 at 11:44
  • 2
    Much depends not only on the fact that the bounds must be respected but also on how the response should behave as the bounds are approached. But beta regression is one approach, as @PeterFlom suggests, and a variant of logit is another. Clearly your response minus 0.25, doubled, has you happily ranging from 0 to 1, and you can reverse the transformation at the end if needed. – Nick Cox Oct 22 '14 at 11:47

0 Answers0