I have a low-level knowledge of statistics. Hoping there can be somebody who can answer my queries. So, I have these dataset and I want to run a multiple linear regression. Sadly my data and the residuals cannot be normalized ever after I transformed my data.

Then, I tried to look for another method to use, I found the Generalized Linear Model. My problem with the GLM is that I cannot identify what "Family" and "Link" to use. I tried to read references and reviews but I cannot find anything in which family data will fall.

My dataset is like this. The dependent variable is continuous but bounded. The values range from 0 to 1 (any number between 0 and 1). My independent variables are a mix of ordinal and ratio variables. I am now looking at quasi family but I am still not sure.

Also, if you know other methods that might help, please send your suggestions.

Thank you in advance!

  • 1
    $\begingroup$ By far the best approach is based on understanding the response variable, because whenever possible you want to select a model that is appropriate for its conditional distribution. What can you tell us about that? $\endgroup$
    – whuber
    Commented Apr 24, 2019 at 14:13

1 Answer 1


For choosing a "type" of regression, the first step is to look at the DV. When the DV is bounded (as yours is) beta regression is often a good alternative to OLS regression.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.