Skip to main content
Cross Validated

You are not logged in. Your edit will be placed in a queue until it is peer reviewed.

We welcome edits that make the post easier to understand and more valuable for readers. Because community members review edits, please try to make the post substantially better than how you found it, for example, by fixing grammar or adding additional resources and hyperlinks.

Required fields*

4
  • $\begingroup$ About logistic regression, I think you're unnecessarily pessimistic. Logistic regression is equivalent to computing posterior class probabilities for a two-class problem with each class Gaussian distributed, with different means and a shared covariance. The MLE for the covariance is just a weighted sum of the per-class covariances, so the sufficient statistics are just the per-class count, sum, and sum of squares. Obviously it's easy to contrive an exact update using the sufficient statistics. $\endgroup$ Commented Apr 22, 2016 at 0:09
  • 3
    $\begingroup$ @RobertDodier You've described linear discriminant analysis, not logistic regression. The last paragraph of the introductory section here clarifies the relationship. $\endgroup$ Commented Sep 3, 2016 at 1:09
  • $\begingroup$ @ahfoss Even if the per-class data are not normally distributed, one can still construct a model equivalent to logistic regression via per-class covariances. $\endgroup$ Commented Sep 3, 2016 at 6:22
  • 1
    $\begingroup$ @RobertDodier What is the equivalent model? You seem to be implying there is an obvious solution to a substantially difficult problem. Your solution is either more brilliant than you realize, or much less so. $\endgroup$ Commented Sep 5, 2016 at 4:36