Proc Means - Customising Output

[Kalai2008]

Hello,

Need to change the proc means output.

Here is the actual output:

proc  means data=test mean median sum;

class countryname ;

var income ;

run;

Country Name	N Obs	Mean	Median	Sum
Algeria	60	$3,676.36	$181.94	$222,764.00
Afghanistan	64	$18,169.66	$6,056.50	$119,997.97

There is another variable called Gender, which has male and female. I need the output to be like this showing only the nobs (counts) instead of displaying statistics for male and female. I dont want to include gender in the class variable to avoid the statistics for the male and female.

Country Name	Male	Female	N Obs	Mean	Median	Sum
Algeria	23	37	60	$3,676.36	$181.94	$222,764.00
Afghanistan	24	40	64	$18,169.66	$6,056.50	$119,997.97

As you can see in this output(Which I need), there is no stats for male and female.

Thanks for looking

[Reeza]

You can't do this via PROC MEANS as far as I know. The easiest alternative is to calculate it externally and merge the data in. 

Depending on how complex your report is, you could generate the table in one step in SQL but it's a fairly manual task. And you would need SAS 9.4+ at least. 

[RW9]

proc  means data=test mean median sum;
  class countryname ;
  var income ;
  output out=inter;
run;
proc sql;
  create table WANT as
  select  COUNTRYNAME,
          (select count(*) from TEST A where A.COUNTRYNAME=COUNTRYNAME and SEX="MALE") as MALE,
          (select count(*) from TEST A where A.COUNTRYNAME=COUNTRYNAME and SEX="FEMALE") as FEMALE,
          NOBS,
          MEAN,
          MEDIAN,
          SUM
   from   INTER;
quit;

[Ksharp]

If you have SAS9.4 . Try SQL.

proc sql;
select age,count(*) as n,
 mean(weight) as mean,
 median(weight) as median,
 sum(weight) as sum,
 (select count(*) from sashelp.class where sex='F' and age=a.age) as Female,
 (select count(*) from sashelp.class where sex='M' and age=a.age) as Male

 from sashelp.class as a
  group by age ;
 quit;

[ballardw]

Provide some data.

I suspect that Proc Tabulate and/or Proc report could do that quite easily.

A complete guess without data:

proc tabulate data = have;

   class country;

   class gender;

   var income;

   table country,

            (gender=' ' All='N obs')*n=' ' income=' '*(mean*f=dollar12.2 median*f=dollar12.2 sum*f=dollar16.2);

run;