4
$\begingroup$

I'm currently working on web advertising topics and I'd like to know if it does make sense to, let's say, compare the resulting weights of two logits?

Let's take as an example the case where input variables are the same across models. A variable gets a weight of 0.22 as the result of the first logit, and a weight of 0.08 as the result of the second logit (all the other variables remaining the same, and their weights may change obviously)

May I conclude that this variable is of highest importance for explaining hidden patterns within the first dataset, or is such a comparison forbidden? (In fact, when thinking about the problem I couldn't find any obvious mistake...)

$\endgroup$

1 Answer 1

1
$\begingroup$

Of course, this makes complete sense, as long as all the other variables are the same. In fact, if you think of logistic regression as a linear regression with the log-odds of the output variable (as in http://www.wright.edu/~thaddeus.tarpey/ES714glm.pdf), then it is trivial that comparing the weights is the same as comparing them in linear regression, which makes perfect sense.

$\endgroup$
1
  • $\begingroup$ awesome, thanks very much for your clear and well-documented answer ! $\endgroup$
    – antounes
    Apr 23, 2018 at 7:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.