topic Re: Mediation for survey design (NHANES) in SAS Programming

Mediation for survey design (NHANES)

maciulei — Fri, 03 Jan 2020 15:56:40 GMT

Hi all,

I working on conducting mediation analysis (linear, logistic, multinomial, and ordinal) using NHANES data that requires inputing survey weights, cluster, and strata. So far, I've been using the code (with macro) from Linda Valeri and Tyler J. VanderWeele, 2013. However, that code doesn't allow me to include the weight, cluster, and strata, and I can only run linear and logistic models. I've been searching for other codes and found the paper by May, Ha, and Soulakova 2019 that can handle survey adjustments, but I am having a hard time learning how to use it, particularly the Fay method that's indicated (code attached). Has anyone dealt with this before or has better suggestions?

Thank you,

P.S. I am using SAS 9.4

Re: Mediation for survey design (NHANES)

katblack — Fri, 03 Jan 2020 20:22:57 GMT

Here's some information that might be helpful for you about the survey analysis features available in SAS https://www.lexjansen.com/nesug/nesug08/sa/sa06.pdf

This provides info for info for logistic and linear regression.

This resource may also be helpful, and some of their examples use NHANES data

https://stats.idre.ucla.edu/sas/seminars/sas-survey/

Re: Mediation for survey design (NHANES)

maciulei — Fri, 03 Jan 2020 20:49:45 GMT

Thanks for the resources, but I am familiar with how to do linear and logistic regressions using the survey specifications. My issue is running a mediation code with macros and adding those survey specifications.