topic PROC SURVEYREG - non normal distribution in SAS Programming

PROC SURVEYREG - non normal distribution

amng24 — Wed, 03 Apr 2024 17:20:33 GMT

Is there a way to use proc surveyreg with a non-normal distribution? My outcome variable is continuous but right-skewed, so my first thought was to use PROC GLM which could relax the assumption of a linear regression to have a normal distribution. However, I have not seen a feasible way to use survey weights with PROC GLM, and I am using complex survey data.

Re: PROC SURVEYREG - non normal distribution

PaigeMiller — Wed, 03 Apr 2024 17:46:12 GMT

Most types of regression do not require normally distributed y-variables. Regression does require normally distributed errors, which are measured by the residuals. If the residuals are not normally distributed, then perhaps you need to traansform the data somehow.

https://blogs.sas.com/content/iml/2018/08/27/on-the-assumptions-and-misconceptions-of-linear-regression.html