I'd like to compare logistic regression models using the BIC. I have 600 data points, but only 30 of these belong to the class of interest. To calculate the BIC I need to know my sample size, $N$.

In my case, is my sample size 600 or 30? Should the BIC calculation use the total size of my dataset (600), or just the number of positive samples (30)?

  • $\begingroup$ BIC, AIC look at all your data. $\endgroup$ – Jon Dec 18 '16 at 4:34
  • $\begingroup$ How many points did you train your classifier on? This is your sample size. $\endgroup$ – quant_dev Jun 21 '18 at 11:56

If you have used all 600 points when fitting your model, then the sample size $n$ is 600.


Your Answer

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

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