regression choice for DV that is continuous between 0 and 2

[Kathi Retti]

Hello!
I have a dependent variable that can take any value between 0 and 2 (0 means no difference, 2 means everything different). Which regression do I use? I was thinking about OLS or Tobit

[Nick Cox]

Tobit sounds wrong unless there could have been values beyond [0, 2].

I'd divide by 2 and use a logit link in a generalized linear model.

Or if your relationships are weak and predicted values outside [0, 2] won't bite you, then OLS might not be utterly bad.

[Kathi Retti]

Thank you Nick!
This is great help. Do you mean weak correlations when you say "relationships are weak"? because I do have weak correlations so that would be handy

[Nick Cox]

I mean low predictability of the outcome or response. The limiting case is no predictability at all, in which circumstance the prediction is the observed mean, which won't be outside the range of the outcome. For low predictability the prediction won't vary much around the observed mean.