Fractional response variables

[deleted_user]

I am looking for a procedure (or similar) in SAS that will analyze fractional response dependent variables, i.e., variables than range continuous from 0 to 1 inclusive. Does PROC GLM handle such?

[Doc_Duke]

Yes and No. GLM can be used when the range of the data is limited. You will often see it used on responses that are 0-100 (like % grades) and that is analogous to your 0-1 range.

However, the assumption of normality is violated and you need to use the Central Limit Theorem to make useful inferences. In practice, this means that the closer your outcomes are to the extreme points in the distribution, the larger your sample size needs to be.

"Large" is a relative term. It would be prudent to do some bootstrap validation of your model to make sure that the parameter estimates are stable.

[deleted_user]

Are you familiar with (or is anyone familiar with) Papke and Wooldridge, 1996, and how they treat fractional response variables?

[mg]

I'm dealing with similar problem .. have you been able to figure it out? What have you used?