- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Hi, can someone answer my naive question?
I am analyzing a cluster randomized trial conducted in schools. My outcome is binary (improvement in 50% of the tooth sites that showed gingival bleeding outcome at baseline after receiving the intervention vs no improvement). The trial has three arms (2 intervention groups and 1 control group). The unit of analysis was the school child.
For my bivariate analysis, I would like to utilize the Chi-square test, to explore if the proportion of students with my outcome of interest differs by group at baseline. I have read that I need to run an adjusted chi square test (adjusted for ICC), but I can't find the SAS code. The clustering is as follows:
1000 students within 19 primary schools within 9 villages. 3 villages were randomized to group 1 (intervention 1), 3 villages to group 2 (intervention 2) and 3 villages to group 3 (control).
I have created the models and ran a GEE (with the appropriate corrections for low number of clusters) for my multivariate analysis that accounted for the clustering. However, I also want to run a bivariate analysis, to include in my statistical analysis section of the paper. I am facing a roadblock because I can't find the SAS code for an adjusted chi-square test that accounts for the clustering within schools (ICC)
.
Thank you!
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
One possibility might be a Rao-Scott design-adjusted chi-square test, which accounts for clustering.
PROC SURVEYFREQ provides this test (CHISQ option in the TABLES statement).
Documentation is here:
and