topic Proc Freq table in SAS Programming

Proc Freq table

Sujithpeta — Fri, 15 Jun 2018 16:01:37 GMT

Hello,

I've a dataset with 10 variables and each takes only 0 or 1 (No/Yes). I want to compare the freq of 'yes' among variables to a specific variables called VLU (also yes/no), shown in the picture.

Thanks in advance!

Re: Proc Freq table

Astounding — Fri, 15 Jun 2018 16:24:44 GMT

A couple of suggestions ...

First, it's easier to use PROC TABULATE to get these results.

Second, I imagine you don't need the frequency of YES for VARIABLE1 through VARIABLE9. Rather, you could use the average value. A side benefit would be that you can condense the table into one row for each of the 9 variables. So, consider this:

proc tabulate data=have;

class VLU;

var variable1-variable9;

tables VLU * mean=' ', variable1-variable9;

run;

There are lots of ways to improve the table appearance, but see if this gets you started in the right direction. You also might need to add a format for VLU, using a format that prints 1 as "Yes" and 0 as "No".

Re: Proc Freq table

ballardw — Fri, 15 Jun 2018 16:26:45 GMT

One way to get started:

proc tabulate data=have;
   class vlu;
   var variable1-variable9;
   table variable1-variable*sum='',
         vlu
   ;
run;

options affecting appearance such as a format for VLU to display Yes/NO instead of 1/0 and such to make this prettier.

Labels for variables also.

Proc freq doesn't let you combine that many variables into a single result, Procs Tabulate and Report would.

Re: Proc Freq table

Sujithpeta — Fri, 15 Jun 2018 16:56:37 GMT

How should I change the code if each variable has different name and not in a sequence to write as Variable1-Variable?

Thanks

Re: Proc Freq table

Astounding — Fri, 15 Jun 2018 17:19:58 GMT

You can replace variable1-variable9 with a list:

variable1 variable2 variable3 ..... variable9

Depending on where it appears in PROC TABULATE, you may need to enclose the list in parentheses.

Also, I think the best table structure would be a combination of what I posted and what @ballardw posted:

tables variable1-variable9, VLU * mean=' ';

You can experiment with using SUM= or MEAN=, to see which you prefer.

Re: Proc Freq table

ballardw — Fri, 15 Jun 2018 19:15:20 GMT

@Astounding wrote:

You can replace variable1-variable9 with a list:

variable1 variable2 variable3 ..... variable9

Depending on where it appears in PROC TABULATE, you may need to enclose the list in parentheses.

Also, I think the best table structure would be a combination of what I posted and what @ballardw posted:

tables variable1-variable9, VLU * mean=' ';

You can experiment with using SUM= or MEAN=, to see which you prefer.

The mean of a 0/1 coded response as @Astounding suggests will give the percentage of 1's. In which case you might want to us

mean*f=percent8.2 or similar. The *f= says to use the format following the = as the format for the resulting statistic.

The sum I suggested give the count of 1's. So use of sum*f=6. would suppress the default decimal portion and assume 6 characters as the longest result.