My dependent variable is number of days in a week a certain activity occurs, so I figured I would express it as a percentage out of 7 (days) and model it using logistic regression.

I would like to know:

  1. Is this the way to model this? I have read mixed things about ratios (here, and here) and want to confirm logistic regression makes sense here, and if not, what would be a better way to model? Linear regression doesn't seem to make sense since the dependent variable (number of days in a week) is bounded between 0 and 7.

  2. If logistic regression is the way to go, how to evaluate goodness of fit? The standard classification metrics don't apply e.g. confusion matrix, pun intended. Intuitively, OLS R^2 would make sense in this case, as I just want to see how close the predicted - in this case continuous variable - is to the actual. I am not sure if the adjusted R^2 is necessary for this type of model.


  • $\begingroup$ Be careful when using logistic regression. It is very likely that over-dispersion/dependency exists, so maybe you need to add random effect in the model. $\endgroup$ – user158565 May 3 '19 at 2:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.