topic Re: Subset dataset to categorical variables only in SAS Programming

Subset dataset to categorical variables only

dbjosiah — Tue, 10 May 2022 18:56:27 GMT

I'm looking for a way to subset a dataset to only its categorical variables (or only continuous, doesn't matter) without knowing the names of the variables. So far haven't found anything that matches what I need. Thanks for any help.

Re: Subset dataset to categorical variables only

PaigeMiller — Tue, 10 May 2022 19:00:08 GMT

Example:

data cars;
    set sashelp.cars(keep=_character_);
run;

Re: Subset dataset to categorical variables only

dbjosiah — Tue, 10 May 2022 19:01:00 GMT

That's so simple it's embarrassing. Thanks.

Re: Subset dataset to categorical variables only

Reeza — Tue, 10 May 2022 19:04:21 GMT

It you can assume categorical variables are character.
For example, cylinders in SASHELP.CARS is a categorical categorical variable but numeric.

Re: Subset dataset to categorical variables only

dbjosiah — Tue, 10 May 2022 21:36:24 GMT

Thanks Reeza. Once I have the real continuous ones out, I'm subsetting the long character ones by hand.

Re: Subset dataset to categorical variables only

Reeza — Tue, 10 May 2022 22:08:00 GMT

If you use SAS Studio, check out the characterize data set task. It does this automatically by looking at the # of distinct values within variables. Anything less than a threshold will be classified as categorical IIRC.

Re: Subset dataset to categorical variables only

dbjosiah — Tue, 10 May 2022 22:09:02 GMT

That would be useful! But I don't use Studio.

Re: Subset dataset to categorical variables only

Tom — Tue, 10 May 2022 22:22:02 GMT

@dbjosiah wrote:
Thanks Reeza. Once I have the real continuous ones out, I'm subsetting the long character ones by hand.

You could try NLEVELS output of PROC FREQ to quickly find the number of levels for each variable.

proc freq nlevels data=have;
  tables _all_ / noprint;
run;

Re: Subset dataset to categorical variables only

dbjosiah — Tue, 10 May 2022 23:40:57 GMT

I've used nlevels, but never like that! Much better than hand subsetting. Thanks!