Conditional logistic regression: Output odds ratios for a continuous variable with a reference

[N1d70]

Hi all,

I am doing a matched (1:10) case-control study looking at the association between income (in ranks) and the risk of stroke.

I use conditional logistic regression:

PROC LOGISTIC desc data = stroke;
strata = matched;
model stroke (event = '1') = INCOMErank;
RUN;

• matched = matching variable
• stroke: 0 = no stroke, 1 = stroke

Now I wish to look at potential non-linearity using cubic splines with knots at the 20, 50, 80th percentile of INCOME rank. Moreover, I wish to output the odds ratio for each rank compared to the 50th percentile INCOME rank. My goal is to achieve a figure with odds ratios on the y-axis and income-ranks on the x-axis.

1. How would you do cubic splines with knots in conditional logistic regression? 
2. How do you obtain odds ratios comparing each rank to a reference for a continuous variable?

Thank you for your time.

[pink_poodle]

1. You can re-use the code from here, combining it with what you have written:
https://communities.sas.com/t5/Statistical-Procedures/How-to-plot-Restricted-Cubic-Spline-in-PROC-LO...
2. "ods output ParameterEstimates=Lgsparms;" line from above link may take care of that.
Also you may try adding the "ods output Logistic.CLoddsPL = o;" before the logistic procedure and add "/orpvalue" option to the model statement. The p-values will be in the "o" dataset.