using a specific matrix as input for cluster analysis

[dustychair]

Hi everyone,

I have a matrix created based on raw data. I want to use this matrix as input for cluster analysis. Can anyone help me to define that matrix as input? Here is the code I use. I tried that code but it did not work. I don't know what I'm missing. I also attached raw data I used.

Thanks 

data tr;
infile 'C:\coll_gen\N100\GROUP10\COMP40\PHI25\gbt\RESULTS-SIG1.OUT';
input CAND_1$ CAND_2$ ID_1 ID_2 THETA_1 THETA_2 MATCHES RT_PROB FLAG ;
run;

proc iml;
use tr nobs nobs;
read all var{CAND_1 CAND_2 RT_PROB};
close;

levels=unique(CAND_1||CAND_2);
matrix1=j(ncol(levels),ncol(levels),.);
mattrib matrix1 r=((levels)) c=((levels)) label='' ;
row=(CAND_1);
col=(CAND_2);
do i=1 to nobs;
r=row[i];c=col[i];p=RT_PROB[i];
matrix1[r,c]=p;
end;
print matrix1;
quit;

ods graphics on;

proc cluster data=matrix1 method=ward ccc pseudo out=tree;
var RT_PROB;
id CAND_1;
run;

ods graphics off;

proc tree data=matrix1 out=New nclusters=5 noprint ;
height _rsq_;
copy RT_PROB ;

run;

[IanWakeling]

All the matrices in IML are destroyed when you exit the procedure with 'quit'.   In order to do what you want, you will need to add the IML statements:

create matrix1 from matrix1;
append from matrix1;

to export the contents of matrix1 to a SAS data set.   There are other problems as matrix1 will be very sparse, mostly with missing values, and you will not be able to run a cluster analysis on it.

[dustychair]

Hi lanWakeling,

Thanks for your response,

I added the statements you mentioned. Now I have the matrix but it does not have the candidates id numbers in rows and columns. I attached the output.

Thanks

[IanWakeling · Accepted solution]

You can export row and column names at the same time as the matrix.  For example:

create matrix1 from matrix1 [rowname=levels colname=levels];
append from matrix1 [rowname=levels ];