Different N obs between proc means and proc npar1way

[Melk]

I am running a 2 sample wilcoxon and also supplementing my p-value with with medians using proc means. However, the number of obs in each are different: proc means reports I have 3 in my class variable whereas npar1way reports I have 2. What can be the reason for this?

proc means data=x median;
    where group = "A";
    class sub;
    var change;
run;

proc npar1way data=x wilcoxon;
    where group = "A";
    class sub;
    var change;
    exact;
run;

[svh · Accepted solution]

You can add 'nmiss' to your PROC MEANS statement to try to examine missingness of the continuous variables at the class level. 

Or, have you tried running a PROC FREQ to examine the categorical variables and their frequencies. That might help.

Such as 

PROC FREQ DATA=X;

  TABLES SUB;

  TABLES GROUP;

  TABLES SUB*GROUP;

RUN;

[Rick_SAS]

In PROC MEANS, "NObs" is the total number of observations in each group. It includes missing and nonmissing. To see the number of observations that are used in the analysis, put N and NMISS on the PROC MEANS statement:

proc means data=x N NMISS median ;
    where group = "A";
    class sub;
    var change;
run;

The "N" column should match the "N" column in NPAR1WAy.