Calculating Percentiles

[bnd]

Hello,

I am trying to calculate the top 25% and bottom 25% for a change score variable. The top 25% indicates high increases in scores and the bottom 25% indicates smaller increases in scores. Is there a way to determine the top 25% of scores and the bottom 25% of scores using proc means or proc univariate in SAS?

Thanks!

[Reeza · Accepted solution]

Use PROC RANK with GROUPS=4.

proc rank data=sashelp.cars groups=4 out=cars;
var mpg_city;
ranks mpg_city_rank;
run;

Haven't verified the code but the idea is there. 

---

@bnd wrote:

Hello,

I am trying to calculate the top 25% and bottom 25% for a change score variable. The top 25% indicates high increases in scores and the bottom 25% indicates smaller increases in scores. Is there a way to determine the top 25% of scores and the bottom 25% of scores using proc means or proc univariate in SAS?

Thanks!

---

[bnd]

Thank you @Reeza! The proc rank procedure allowed me to group the variable into quartiles. But, how would I go about getting the actual top 25% of scores and the bottom 25% of scores once the scores are grouped into the four quartiles? Also, the groups start at 0 rather than 1. 

[Reeza]

Group=0 are your bottom 25% and Group=3 is your top 25%. Yes the groups start from 0 and go to 3, you can add one to it, if that helps you.

[bnd]

Okay, now I understand! Thanks so much! @Reeza 

[PaigeMiller]

---

@bnd wrote:

But, how would I go about getting the actual top 25% of scores and the bottom 25% of scores once the scores are grouped into the four quartiles?

---

PROC UNIVARIATE, PROC MEANS or PROC SUMMARY will calculate these percentiles.

[bnd]

Thanks so much! @PaigeMiller 

[ballardw]

The report procedures Tabulate and Report as well will do some percentiles. An example of calculating P25 and P75 for 3 variables, horsepower mpg_city and weight grouped by 3 categories, type drivetrain and number of cylinders from the SASHELP.CARS data.

proc tabulate data=sashelp.cars;
   class type  drivetrain cylinders;
   var horsepower mpg_city weight;
   table type  drivetrain cylinders,
         (horsepower mpg_city) *(p25 p75)
   ;
run;

[PaigeMiller]

PROC UNIVARIATE, PROC MEANS and PROC SUMMARY will compute the 25th percentile, the 75th percentile and many other percentiles upon request.