On path correlation and PERT bias