count missings for all character variable

[Xiaoyi]

Dear SAS experts, 

I want to just count missing frequency for all character variables by a categorical var (Year) in a dataset.  and have a output dataset. I only care about the missing count, so leave the not missing out. and I only want the character variables, so don't try to conquer all var in one shot and expect perfect results. (I've searched and tried, hasn't found one), but you do have a perfect solution that can do it all (by category and with results output), I think everyone would love to have it :). Keep it separate and keep it simple. Thanks.

The following is a simple code to do the numeric variable FYI. I want similar results for character variables: 

proc means data=sensit_dist nway noprint;
class infyob_yot;
output out=misschk n= nmiss= /autoname;
run;

[ballardw · Accepted solution]

Something like this may get you started. Untested as we don't have your data.

The Value in the data step takes 1/0 values when the variable is missing for the named variable. This creates one record for each variable. The Proc summary/means sums the variable to get a total count of missing.

data temp;
   set sensit_dist;
   array c _character_;
   length name $ 32;
   do i=1 to dim (c);
      name = vname(c[i]);
      value = (missing( c[i]));
      output;
   end;
   keep  infyob_yot name value ;
run;

proc summary data=temp nway;
   class infyob_yot name;
   var value;
   output out=work.charmissing Sum= ;
run;

[Xiaoyi]

@ballardw  Thanks for the code this look promising, but it only did for one variable. I had 37 character variables. I tried to tweak it a little bit to make it work, but I run it, and every time, it only counted one variables. It didn't give me a dataset with 37 variables in it. You can use any datasets you have to test it. 

[ballardw]

---

@Xiaoyi wrote:

@ballardw  Thanks for the code this look promising, but it only did for one variable. I had 37 character variables. I tried to tweak it a little bit to make it work, but I run it, and every time, it only counted one variables. It didn't give me a dataset with 37 variables in it. You can use any datasets you have to test it. 

---

Check the post with the code for the added OUTPUT statement.

The problem with not having a data set is code cannot be tested...