I need to establish the minimal detectable change for an outcome to distinguish true change from test-retest variability. The data has 30 eyes of 20 participants with 3 scans on the same day. A grid of 25 squares (boxes) is applied to each of the 3 scans and the outcome is obtained in each box for a total of 30×3×25 = 2250 observations. In the literature, minimal detectable change is defined as
MDC95 = sqrt(2)×1.96×SEM
where SEM is standard error of measurement and is the same as within subject (residual) standard deviation.
I am using the following SAS code
proc mixed data = dat method=REML;
class StudyID Eye box;
model y = / solution ddfm=kr;
random intercept / subject=StudyID;
random intercept / subject=Eye(StudyID);
random intercept / subject=box(Eye StudyID);
I think I have approached this correctly but I would definitely appreciate confirmation.
It's been awhile since I've done a model like this. You have the first step correct, or at least workable. You need add predictors to your y = model, and it may also be necessary to add random slopes for your "boxes", depending on model fit.
Registration is open! SAS is returning to Vegas for an AI and analytics experience like no other! Whether you're an executive, manager, end user or SAS partner, SAS Innovate is designed for everyone on your team. Register for just $495 by 12/31/2023.
If you are interested in speaking, there is still time to submit a session idea. More details are posted on the website.