I wouldn't worry to much about it, logistic regression is a much better method and has optimality properties. Best wishes David Booth PS see the reference below

Hi there,

I would second David's suggestion. In their (free!) book (p. 161) introduction to statistical learning, James et al. compared several methods for binary classification, and logistic regression was superrior unless there were interactions (for which quadratic LD was better) but the authors note that this could be achieved by including polynomials.

You can get the book here:

https://www.statlearning.com/

HTH

Holger