topic Re: How to robustly scale variables? in SAS Programming

How to robustly scale variables?

EpiMoby — Tue, 04 Apr 2023 18:59:29 GMT

I have a set of county-level indicators that I am trying to robustly scale for comparability across regression models. I've tried many things but can't seem to find the code to do this.

Re: How to robustly scale variables?

PaigeMiller — Tue, 04 Apr 2023 19:03:36 GMT

Please explain what you mean by "robustly scale for comparability across regression models". Why would scaling even matter in Regression? If you have variables temperature and distance, it doesn't matter if temperature is Fahrenheit or Celsius; and it doesn't matter if distance is kilometers or 1/16 of an inch, you will get the same predicted values and the same goodness of fit either way.

Re: How to robustly scale variables?

EpiMoby — Tue, 04 Apr 2023 19:21:50 GMT

For comparability across regression models, I need to robustly scale each variable by centering
at the median value among all counties and then standardized to the median of the absolute deviation of
all counties. I'm doing this so 1 standard deviation means the same for all variables. I can do this in R, but am trying to replicate in SAS. The code I'm using is

proc stdize data = dataset out = scalesample method=median;

var variable;

run;

Re: How to robustly scale variables?

PaigeMiller — Tue, 04 Apr 2023 19:32:01 GMT

Seems like you want to use PROC STDIZE with the option METHOD=MAD

https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.4/statug/statug_stdize_syntax01.htm#statug.stdize.methodopt