Suppose that you have a regression model,

y=x*beta+error;

then beta=(x'x)**(-1)*(x'y); (*) under regression model assumptions which you can find in any regression textbook.

The beta calculation depends on the invertible moment matrix (x'x). If x'x is singular, then x'x can not be inverted.

When x'x is singular, it usually means that there is a linear combination among the coefs. you may over parametrize the model.

Your model may involve weighted regression so that moment matrix is in the form of x'wx

Here is an example. HTH.

data t1; do i = 1 to 30; x1=rannor(123); x2=x1; y=3+3*x1+3*x2+rannor(123); output; end; run;

proc reg data=t1; model y=x1 x2; run; quit;