0
$\begingroup$

In logistic regression, logit Y = ax1+ bx2 +cx3. a variable has a coefficient attached to it. So we can directly measure the change in y per unit change in x. Can I get a similar relationship from odds ratio? per unit change in x contributing to how much change in y. Basically I want to understand the contribution to score. E.g. a variable has odds of 20:1 compared to 6:1

$\endgroup$
0
$\begingroup$

Your first line is incorrect. In a logistic regression, Y is not linear in the parameters. Indeed, that's one reason why we need logistic regression. The logit of y is linear in the parameters.

The parameters output from a logistic regression lack an intuitive explanation, unless your intuition about equations with exponentials in both numerator and denominator are a whole lot better than mine.

The odds ratio does what you think the parameter does. That is, the OR measures how the odds of Y change per unit increase in each X.

$\endgroup$
  • $\begingroup$ sorry.. i meant logit of y. $\endgroup$ – Amit Agarwal Feb 11 '16 at 12:41
  • $\begingroup$ sorry.. i meant logit of y. my equation is a linear regression. say i had 10 variables in the equation. the top one had an odds ratio of 20:1 while the lowest one is 6:1. But I intuitively feel that the lowest one is more valid and I have only got these odds due to lack of training data. Should I favor the top variable compared to the lower. I just want to understand how much difference can result in output. something similar to linear regression. $\endgroup$ – Amit Agarwal Feb 11 '16 at 12:51
  • $\begingroup$ I don't think this is at all the way to build models. $\endgroup$ – Peter Flom Feb 11 '16 at 18:06
  • $\begingroup$ if you have quality data then definitely not, however if you have data challenges, especially in text mining, i think you have to be on the lookout for potential glitches and trust common sense rather than blindly following a model. $\endgroup$ – Amit Agarwal Feb 12 '16 at 2:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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