Thanks a lot ksharp for the solution. It was extremely helpful.
I have modified the code a bit as per my requirement as below. I am facing 1 problem i.e.
If i have certain number of blanks in P_Id B which means P_ID A has 9 classes whereas P_ID B has only 5, so the last 4 columns has ". "values. How to exclude these column from the do loop as it gives an error. I tried supressing them to 0 but calculations didnt work because number of rows are 5 and columns become 9.
I think there is some filter required in the do loop while assigning the values to variable c so that it can ignore all dots"." related column and matrix multiplication should work then.
Thanks for your help again.
data have;
set wide1;
run;
proc iml;
vnames=contents(have);
var_c=vnames[loc(prxmatch('/^C\d+\s*$/',vnames))];
use have;
read all var{p_id};
y="Y0":"Y20";
id=unique(p_id);
do i=1 to ncol(id);
read all var {year} where (p_id=(id[i]));
read all var var_c where (p_id=(id[i])) into c;
do j=1 to 20;
if(j=1) then
do;
idx_y1=loc(year=(y[j]));
y1=c[idx_y1,];
end;
else if (j<5) then;
do;
idx_y1=loc(year=(y[j-1]));
idx_y2=loc(year=(y[j+1-1]));
y1=c[idx_y1,];
y2=c[idx_y2,];
end;
else;
do;
idx_y1=loc(year=(y[4]));
idx_y2=loc(year=(y[4]));
y1=c[idx_y1,];
y2=c[idx_y2,];
end;
if j= 1 then;
do;
yxy=y1;
want=yxy[1:(ncol(yxy)),ncol(yxy)];
end;
else if j=2 then ;
do;
want=y1[,1:(ncol(yxy)-1)]*y2[1:(ncol(yxy)-1),ncol(yxy)];
yxy= y1 *y2;
end;
else if j>2 then;
do;
want=yxy[,1:(ncol(yxy)-1)]*y1[1:(ncol(yxy)-1),ncol(yxy)];
yxy= yxy *y1;
end;
/* print "Result for" (id[i]) (y[j])"-"(y[j+1]),yxy,want;*/
/* label=id[i];*/
/* label1=y[j+1];*/
label=repeat(id[i],nrow(want));
x=repeat(y[j+1-1],nrow(want));
labels=labels//label;
label1=label1//x;
wants=wants//want;
end;
end;
close;
create Year var {label1};
append;
close;
create P_id var {labels};
append;
close;
create value from wants;
append from wants;
close;
quit;
data want;
merge P_id Year value;
run;
