What is your recommended analysis for the following problem:
I have 3 (treatment) groups and a non-randomized pre-post design. I want analyze the effects of the groups on the follow-up measure, controlled for the baseline measure.
If the pre-post measures were continuous, I would go with ANCOVA, using the follow-up measure as the dependend variable, the group variable as the factor and the baseline measure as a covariate. But what is your recommended analysis for binary pre- and post- measures?
* 3 treatment groups
* binary outcome (follow up)
* controlled for a binary covariate (baseline)
Thank you in advance for your help .
Hi Christian,
In this situation, conducting a logistic mixed-effects regression including 'Group' (3 levels), and 'Time' with two levels (i.e., time1, time2) is the only approach I can think of. It will be important to give 'Subject' a random intercept in the model as well since the same individuals are being measured twice (1|Subject).
If you're familiar with R software, you can conduct such an analysis using the lmer() function. Something along these lines:
model
Dear Christian Grebe ,
I think Blaine Tomkins has the right idea. This is a binary logistic regression. Your treatment group needs to be replaced by two dummy variables. There may be an interaction between these and your baseline variable.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA