topic Contrast for mixed model with spline and one continuous predictor with interaction in SAS Procedures

Contrast for mixed model with spline and one continuous predictor with interaction

louise0828 — Fri, 20 Oct 2023 22:22:01 GMT

I have a dataset with a continuous variable specifying the cumulative dose of a drug received at time t (drug received at multiple time points), a visit day variable, and a continuous outcome variable. I am fitting a model that cubic spline model with knots at the study days the drug is received to model the outcome variable with cumulative dose and a spline for visit day as predictors (and the interaction of day with dose) with subject specific intercept and slope. Below is the model code that I've used:

proc glimmix data=dat;

class subjectid;

effect = spl(spline day/degree=3 knotmethod=list(5 10 15));

model outcome = spl | dose;

random intercept day / subject=subjectid;

run;

I would like to estimate the contrast for a subject receiving a dose of 20 and a subject receiving a dose of 10 at day 12. I tried to write the following estimate statement:

estimate "Dose 20 vs Dose 10 at Day 12" spl*dose [1, 12 20][-1, 12 10];

But the estimate is showing as "Non-est". Am I writing the contrast wrong?