1

Why we do not use R squared for logistic regression? What is the logic behind it?

Richard Hardy
  • 67,272
Aslı
  • 11
  • A more general concept is deviance.

    In the case of Gaussian distributed errors the deviance is equal to sum of squared error. That sun of squared error and the related R-squared is a special case.

     – Sextus Empiricus Apr 06 '22 at 19:09
  • can you be more spesific? Basicly how can we prove that r squared is not valid for logistic regression? – Aslı Apr 06 '22 at 19:14
  • There is no logic behind it in the sense of a rigorous proof. That is because R^2 neither has proof when it is used for an ordinary linear model. R^2 is just a Pearson correlation between observed values and modeled/fitted values. Correlations with variables that have only values 0 and 1 can be done as well. It doesn't need proof. It is just that correlations with normal distributed variables have a more intuitive explanation. – Sextus Empiricus Apr 06 '22 at 20:00

0 Answers0