topic Re: Variable Transformation in SAS Forecasting and Econometrics

Variable Transformation

ash3 — Fri, 15 Jun 2018 12:29:52 GMT

I have 50 continuous variables and I want to check normality condition of each variable. Is there any syntax available which suggest me automatically best transformation (whether it is log, exp or sqrt) of each variable in one go.

Re: Variable Transformation

ballardw — Fri, 15 Jun 2018 14:58:59 GMT

I would start with Proc univariate as you can request a test for normality as well as get some descriptive statistics and/or plots of distributions that would give a hint which transform might be useful for those that aren't normal, or close enough for your purpose.

Proc univariate data=have normal;

var <list your variable here or use on of the SAS variable list forms>;

run;

would get you started. Histogram or CDFPLOT or other of the plot statements might be helpful.

Re: Variable Transformation

haoduonge — Wed, 28 Sep 2022 19:41:08 GMT

Hi all,

I want to continue with this question.

In STATA we have qladder to check that, I hope SAS has something similar by now too.

Hao

Re: Variable Transformation

Ksharp — Thu, 29 Sep 2022 11:54:29 GMT

1) Try Box-Cox transformation, @Rick_SAS wrote a blog about this:
https://blogs.sas.com/content/iml/2022/08/22/box-cox-transform.html
https://blogs.sas.com/content/iml/2022/08/17/box-cox-regression.html

2) Check PROC GENMOD 's ASSESS Statement

Re: Variable Transformation

PaigeMiller — Thu, 29 Sep 2022 12:01:59 GMT

@haoduonge wrote:

Hi all,

I want to continue with this question.

In STATA we have qladder to check that, I hope SAS has something similar by now too.

Hao

we don't know what qladder does, please explain

Also, what is wrong with doing what @ballardw said above? Seems to be a rather complete answer to the original question.

Re: Variable Transformation

haoduonge — Sun, 02 Oct 2022 21:06:00 GMT

I meant the code that can be used to exam multiple variables silmultaneously (avoid to run over and over again).

Thanks!

Hao