I've done a presentation recently where I'll be using a binary logistic regression. However, one of the audience said that I should be using a conditional logistic regression.
Can anyone tell me whats the difference? I thought it is the same.
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Perhaps a bit late to answer this question. Binary logistic (BL) and condition logit (CL) regressions are the same thing depending on model specification. My understanding is that BL is used to model a binary (0/1) event (e.g., would you buy this product? Yes/No), while CL is used to model a "pairwise" event (e.g., Which product would you buy? A/B). In the particular case of a "pairwise" event instead of modelling the decision between A/B, you could model the difference between the 2 events (A - B) as a binary event (0/1) (e.g., Is the difference between (A-B) big enough to buy A? Yes/No). You would obtain exactly the same results. I guess that's why people sometimes use BL and CL (or even multinomial logit (MNL)) interchangeably.
