I am calculating the odd ratio of logistic regression (using statsmodel of Python). I have one independent variable (i.e. process type: faulty (1) or non-faulty (2) and one dependent variable (i.e. process-time: late (0) or on-time (1)). I calculated the odd ratio at C.I 95% using logistic regression (I used statsmodel of Python). I got the following value:

                    5%        95%      Odds Ratio
Process type    1.431001    1.541844    1.485389

I interpreted the above odds ratio as "Increasing from 1 to 2 for Process type (i.e. going from a faulty to non-faulty) is associated with an increased in the odds by 48% of completing the process on time."

My first question: is my interpretation correct?

As the independent variable is nominal in nature, do I need to calculate the separate odd ratio for each value type of independent variable (i.e. one odd ratio for faulty values and one add ratio of non-faulty values)? My understanding is that as I am using logistic regression to calculate odds ratio so I should be using an independent variable that has at least two types of nominal variable values. I can not calculate for a single type of nominal value.

Could anyone shed some light on it?

  • $\begingroup$ just confirming, non-faulty is your control group? $\endgroup$
    – jros
    Jul 22, 2021 at 18:01
  • $\begingroup$ Non-faulty is one of the value type of independent variable. What do you mean by control group??? $\endgroup$ Jul 22, 2021 at 19:26
  • $\begingroup$ For the interpretation, are you comparing faulty to non-faulty or non-faulty to faulty? $\endgroup$
    – jros
    Jul 22, 2021 at 19:32
  • $\begingroup$ @jros non-faulty to faulty $\endgroup$ Jul 22, 2021 at 20:00

1 Answer 1


The interpretation should be adjusted slightly to something more similar to: The odds of a non-faulty process experiencing an on-time process-time are 48% greater than the odds of a faulty process experiencing an on-time process-time.

In order to calculate the odds ratio for faulty process, just flip the odds ratio (1/1.485).

The coefficients of a logistic regression rely on comparison, so even continuous, numerical independent variables have an comparative interpretation


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.