I would like to explain a continuous probability, beta distributed (U shaped) and continuous on [0,1], through a glm model with a logistic link function.
Are there any concerns I should take care of ?
Asked
Active
Viewed 23 times
0
kjetil b halvorsen
- 63,378
- 26
- 142
- 467
user19727
- 41
- 1
- 2
-
1Beta regression (which itself is not a glm) would be a natural choice but it only works for dependent variables in $(0,1)$. So if the dependent variable has zeroes or ones, beta regression won't work. – COOLSerdash Oct 25 '21 at 09:56
-
Thank you very much, yes the variable is bounded between 0 and 1 but can takes any values. There are values equal to one but this is pretty rare. In my case, the logisitc function is the natural choice because extreme values of my regressors are linked to high and low probabilities. – user19727 Oct 25 '21 at 11:33
-
Have a look at https://stats.stackexchange.com/questions/216122/what-is-the-difference-between-logistic-regression-and-fractional-response-regre – kjetil b halvorsen Oct 26 '21 at 01:47