I came across following definition of crude odds ratio in binary logistic regression:

Crude odds ratio is when you have only one independent variable

And also I heard:

Crude odds ratio is when you have only one binary independent variable (2x2 table)

Which is proper one?

  • $\begingroup$ Well, to have an independent variable implies you must also have a dependent variable (binary) as well as a model (binary logistic regression). Both of your definitions omit this important information, so I would strike them off on those grounds. $\endgroup$ Jun 17, 2018 at 17:41
  • $\begingroup$ I think a better terminology for the odds ratio you invoke is "unadjusted" (rather than "crude") - for that definition to make sense, you must also have other independent variables and/or covariates you are not adjusting for in your model. $\endgroup$ Jun 17, 2018 at 17:45
  • 1
    $\begingroup$ Hi, Isabella thanks for noticing lack of "binary", I corrected that. $\endgroup$
    – mokebe
    Jun 17, 2018 at 19:07

1 Answer 1


The former definition is correct. Logistic regression gives us a notion of odds ratios for continuous regressors. In the past, we preferred binary regressors because the models could be calculated without technology. Now we have a broader perspective because of our computational power.

  • $\begingroup$ Could you give paper or book or any other reliable source, which confirms this? $\endgroup$
    – mokebe
    Jun 17, 2018 at 15:52
  • $\begingroup$ @mokebe I believe modern epidemiology by Rothman has such a discussion. The term is not a mathematical one, but rather one of causal modeling and epidemiology. $\endgroup$
    – AdamO
    Jun 17, 2018 at 19:16

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.