Create a nullvector to have a nxn-matrix to calculate MWB

[Rick_SAS]

> do I have to have it? 

I included that information so that you can easily check the names of the variables in the input matrix. But you can ignore that suggestion if you prefer an alternative. 

[Tom]

Before of course.  Fix the data using normal coding before venturing off into the world of matrix operations.

[Tom]

Fix it before you get into IML.

You have 7 observations and only 6 analysis variables.  If you want to make sure there are 7 analysis variables then just make a new variable that is all missing.

In your case you seem to be naming the variable MK1 to MK7 but you didn't create the MK1 variable.

So you can just run this code.

data square;
  length scale $8 mk1-mk7 8 ;
  set psuedo_data;
run;

If you don't know which how many MK variables there should be just first count the number of observations and put it into a macro variable.

proc sql noprint;
select count(*) into :nobs trimmed from psuedo_data;
quit;
data square;
  length scale $8 mk1-mk&nobs 8;
  set psuedo_data;
run;

[cleokatt]

yes, that was the problem that my matrix was nx(n-1)