Create condition on one dataset based on another dataset

[mariko5797]

I have two datasets that look like the following:

I want to perform ANOVA only if there are more than 3 subjects in each BMI category for each time period (STPER, ENDPER). In my head it would be something along the lines of

%if have.STPER > 3 (where BMI = 'Not Obese' and STAT = 'N') and have.STPER > 3 (where BMI = 'Not Obese' and STAT = 'N') and have.STPER > 3 (where BMI = 'Not Obese' and STAT = 'N') %then %do;
 proc glm data = have2;
  class BMI;
  model log(CONCEN) = BMI;
  means BMI / hovtest = levene welch;
  ods output welch = welch diff = pairs;
 run;
/*repeat for ENDPER*/

I have very little experience with macros, so I'm not sure if what I am asking is possible.

[mkeintz · Accepted solution]

You don't need full-blown macro code here.  Instead you can use the "select ... into :macrovar" facility of PROC SQL to populate a the macrovar &BMI_LIST with a comma separated list of quoted BMI levels that satisfy your requirements:

proc sql noprint;
  select quote(strip(bmi)) into :bmi_list separated by ','
  from have
  where stat='N' and stper>3;
quit;
%put &=bmi_list;

Then run your PROC GLM code, with the additional where statement:

proc glm data = have2;
  where bmi in (&bmi_list);
  class BMI;
  model log(CONCEN) = BMI;
  means BMI / hovtest = levene welch;
  ods output welch = welch diff = pairs;
 run;

Of course, you'd have to run this twice, once for STPER and once for ENDPER.

[ballardw]

Since your Have set has multiple statistics perhaps you only meant the rows where the N statistic matters?

Combine the Have and Have2 so that there is a NEW variable that has the N. Then use a Where in the model to select appropriate records.

Not tested code for obvious reasons:

Proc sort data=have2;
   by bmi;
run;
Proc sort data=have (where=(stat='N')) out=bmin;
   by bmi;
run;

data want;
   merge have2  
              bmin (rename=(stper=stper_n))
   ;
   by bmi;
run; 

Proc glm data=want;
   where stper_n > 3;
   class BMI;
<rest of glm code goes here>

You might save some minor coding by creating a summary set with only the count and use that instead of Have with all the statistics.