I have one continuous variable GENETIC SCORE and one binary variable HEALTH (case vs control i.e. 0/1). I want to fit a log regression model to this data to get the odds ratio for different percentile categories of GENETIC SCORE variable. One way to do this is to simply use a continuous variable. However I want to get the result in the form of

Percentile (%) - OR
<1 - 0.30
1-5 - 0.40
5-10 - 0.49
10-20 - 0.59
20-40 - 0.77
40-60 - 1.00
60-80 - 1.29
80-90 - 1.65
90-95 - 2.03
95-99 - 2.52
.>99 - 3.60

Notice the baseline category 40-60 with OR = 1. So, my first guess (after a long research) is that GENETIC SCORE is categorized in 11 percentile groups. I am aware of all disadvantages of binning. But still, I want to get the results in this form.

My main question here is: how do I setup the model with the desired baseline group? In other words; how do I setup the model in the way that OR is 1 for the middle percentile group (either with continuous or with categorized variable).


To present the results in that form you do not need to represent the data in that form! Just use the data as is, numerically, and then you can represent the results in that form, as a table.

  • $\begingroup$ Yes, I agree. But still - I am asking how to make a transformation from the log regression output (with cont. variable) to this kind of table? $\endgroup$ – user3687798 Dec 4 at 7:17

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.