07-20-2005, 12:40 PM
Hi, I'm trying to code the solution of a real symmetrical eigensystem in fortran. First I used ISML but the results were not satisfactory. Then I tried the NR subroutine 'jacobi' and I'm still getting bad results.

I know that results are not correct by comparing my fortran output with the output obtained by repeating the exact same steps using MATLAB (routine eig). When solving for the principal components of a dataset, the condition P'P = I is satisfied in MATLAB, but not with the fortran results.

I'll be grateful for any help anyone can provide.