topic Re: Help needed with excluding low frequency variables from data set in SAS Procedures

Help needed with excluding low frequency variables from data set

lupefine — Fri, 07 Feb 2014 13:45:20 GMT

Ok, bare with me here, I have only been using SAS for 2 weeks, so I realise some of my sentences will sound like soeone tying to learn a new language.

But here goes.

I have gathered data from our kennel club, one data set containing veterinary information (if the dog has a specific disease (PL= patellar luxation), and in that case, which degree, 0-3), and another 14 for each one of 14 breeds, containing names, registration numbers, lineage ect ect. I ended up with a dataset contaiing 250 000 observations, divided into 14 variables.

I have merged these files by breed code, registratio number, birth date and sex.

I have done some other ifs and formats to get what I want. So far so good.

I proceeded to see how many examinations were done by each veterinarian (variable called clinic). Also I wanted to see the frequencies of degrees (variable = degree) between each veterinarian.

So I wrote the following;

Proc sort data=plallbreeds;

by clinic degree;

PROC FREQ Data=plallbreeds ORDER=FREQ;

Tables Degree;

By Clinic;

Title 'bla bla bla';

Run;

And again, so far so good! Only problems is now that I have 390 veterinarians in my end result, and many of these have done less then 10 exams.

I would like to be able to exclude these from my output so I can better overview the result.

I have searched and searched but I just cant get the code right to do this.

Can someone here help me?

Re: Help needed with excluding low frequency variables from data set

LinusH — Fri, 07 Feb 2014 14:23:18 GMT

If you just want the basic frequency count, you could use SQL:

proc sql;

select clinic, degree, count(*) as Freq

from plallbreeds

group by clinic, degree

order by clinic, degree;

quit;

Re: Help needed with excluding low frequency variables from data set

Astounding — Fri, 07 Feb 2014 15:51:15 GMT

If you are happy with what you have so far, except that you want to eliminate the low counts, here is a way to modify the approach. Instead of printing the report with PROC FREQ, create a data set holding the results. The modification would be:

tables degree / noprint out=clinic_counts;

Then use PROC PRINT to print selected observations. For example:

proc print data=clinic_counts;

where count >= 10;

*by, var, title statements as appropriate;

run;

Good luck.