- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
I am doing a longitudinal analysis using Population Assessment of Tobacco and Health (PATH) data, which uses 1 final weight and 100 replicate weights and Fay's method of balanced repeated replication (BRR) to account for complex survey design (example code shown in image below).
I'm not familiar with this sort of survey design, and this is my first time working with longitudinal data.
Does anyone know if there is a PROC in SAS which will allow me to incorporate both VARMETHOD, WEIGHT, and REPWEIGHTS statements in conjunction with a REPEATED statement? Or does anyone know of some other way to implement all this?
Thanks.
Accepted Solutions
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
There is no way to conduct this BRR method in a mixed effects model in SAS. You can manually achieve this BRR method by following these instructions from the PATH user guide:
"Although software packages do not universally accommodate replicate weights for all analytic methods, the replication method can be applied by repetition to any analytic routine. That is, the desired analysis would first be run using the full-sample weight. Then, it would be repeated replacing the full-sample weight by each replicate weight, in turn (i.e., 100 times for PATH). The formula for BRR variance estimates (provided in equation 5.2.1) would then be used to estimate the variance of any parameter (e.g., regression coefficient) of interest."
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
There is no way to conduct this BRR method in a mixed effects model in SAS. You can manually achieve this BRR method by following these instructions from the PATH user guide:
"Although software packages do not universally accommodate replicate weights for all analytic methods, the replication method can be applied by repetition to any analytic routine. That is, the desired analysis would first be run using the full-sample weight. Then, it would be repeated replacing the full-sample weight by each replicate weight, in turn (i.e., 100 times for PATH). The formula for BRR variance estimates (provided in equation 5.2.1) would then be used to estimate the variance of any parameter (e.g., regression coefficient) of interest."