I'm no expert on the different types of algorithms to compute eigenvalues and vectors for a real, symmetric matrix (coming from linear mass and stiffness matrices for a frame FEA model). I am looking for a good blend of computational efficiency and ease of coding for a system of ~300 DoFs where I only need the first N (N between 1 and 10) eigenvalues and vectors. I will be programming it in VBA, so the less bundled algorithms (e.g. decompositions and things) that I need to program are a factor, too.
I was figuring on inverse power iterations with shifting, but am concerned about stability or skipping an eigenvalue if I shift too much or too little. Is this a realistic concern?