Dear, everyone. Recently, i want to use logistic regression to solve the relationship between a binary character and elevation. When i use logistic regression without quadratic term, its AIC value always low that regression with quadratic term. But some guys told me that adding a quadratic term in logistic regression is problematic. Is there anyone can tell me whether adding quadratic term is feasible or problematic in a logistic regression?
Well, a lower AIC value actually denotes a better fit.
A fundamental problem is interpretation. Unless your goal is simply to predict future events without seeking any theoretical explanations, you'll have a hard time explaining what elevation2 means. In particular, linear and generalized linear models commonly assume a monotonous relationship between predictor and response, which is no longer the case with the hyperbole a quadratic term creates.
Thank you, Martin Schmettow. can you use logistic regression with a quadratic term to plot the trend of a binary character varing with elevation？because the actual trend is a parabolic curve, if i use a logistic regression without a quadratic term, the trend will become a monotonic decreasing. or is there other methods can do this? Best
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression