I am having difficulty getting results when running these modules for multiple regression. I have set up sample problems for 1, 2 and more variables by generating M "slightly" random results for a known linear equation (b= a1*x1 + a2*x2+...aN*xN). The results show INFO=0, RANK=1 and only give one output value in B. When the problem is setup for N=1, the correct solution is given and when N>1, the single solution would still be correct if only the a1 set of data where considered.

The call is shown below:

call SGELSD(ndfv,nv, 1,sgela, ndfv, sgelb, ldb, sgels, RCOND, iRANK, sgelwrk, LWORK, IWORK, INFO )

where ndfv=M=50, the number of sets of data, nv=N, the number of variables, NRHS=1, sgela(M,N) input, sgelb(M,1) results matrix, RCOND=0.0001(,or -1.)

I have also tried SGELSS with the same set of input and results.

I am compiling gfortran on a MacBookPro and have successfully used Lapack module SSYEV in the same program.

Thank you,

Ed