Max/Min of character column with proc means

[tfarkas]

How can I get the maximum of a character field using proc means? I get "ERROR: Variable TERM_CODE in list does not match type prescribed for this list", with the following code.

proc means data=stulasr.vday_summary_eg max;

var TERM_CODE;

run;

I would find it surprising if proc means were unable to generate order statistics for a alphanumeric sample. proc sql works.

Ultimately, I'm trying to create a character macro variable for downstream use in a proc sql step. Perhaps there's a better approach in general. 

[PaigeMiller]

Max/Min of a character column is undefined.

PROC MEANS will not operate on character variables.

Can you give us a small example of the data you have and the output you want?

[tfarkas]

Thanks. Edited my post.

[PaigeMiller]

Ultimately, I'm trying to create a character macro variable for downstream use in a proc sql step. Perhaps there's a better approach in general. 

I don't think there's enough information in these two sentences for anybody to help.

Again, I ask for a small example of your data, and the desired output. (And put it in a reply at the bottom of this thread instead of at the top of the thread).

[novinosrin]

Hi @tfarkas   

"How can I get the maximum of a character field using proc means?"

The idea or the design that you are trying to flow downstream in itself is not the best in my humble opinion and I am pretty certain most people would mostly likely will advice to redesign your flow that accommodates robust approaches in SAS.

Yes, I agree proc sql works fine and so will datastep like

data w;
length want $1;
do char='c','b','a';
want=char<>want;
end;
run;

However the analysis variables in sas procs doesn't quite do the collating sequence of sorts be it EBCIDIC or ASCII. 

In any event, I am sure the pool of experts would be pleased to offer you a clean solution if you can post your requirement clearly.

[Tom · Accepted solution]

PROC MEANS is optimized for generating statistics on numeric variables. You can use character variables as CLASS variables with PROC MEANS.  You could use that to get counts:

proc summary data=have nway;
  class term_code;
  output out=want ;
run;
data _null_;
  set want end=eof;
  if _n_=1 then put 'MIN ' term_code=;
  if eof then put 'MAX ' term_code=;
run;

If you just want min and max of a character variable then PROC SQL is a good choice.

create table min_max as 
select min(term_code) as min_term_code
    , max(term_code) as max_term_code
from have
;

If you want the full distribution of values use PROC FREQ.   Or perhaps you want PROC RANK?