Things to Notice:

Everything but the analytic solution sucks rocks for small separation
Everything is pretty much equally as good for large separation
Monte Carlo methods especially stink for small separation, since samples are likely to have large theta
Regular sampling also stinks at small separation due to theta being large for most samples
Of these approaches, approximating a complex shape by a bunch of disks whose size is on the order of the separation yields the best results when an analytic solution is not available.

Source	Distance=0.05, insensity scale=1.0	Distance=0.25, insensity scale=1.0	Distance=1.00, insensity scale=3.0	Distance=4.00, insensity scale=20.0
Problem 1 1x1 square source uniform (random) sampling dA-to-dA form factor 16 samples
Problem 2 1x1 square source regular sampling dA-to-dA form factor 16 samples
Problem 3 source is 16 small disks regular sampling dA-to-disk form factor 16 samples (square source approximated as 16 disks of area 1/16)
Problem 4 1x1 square source analytic solution dA-to-poly form factor (Stokes' Theorem)
Not Assigned 1x1 square source jittered (random) sampling dA-to-dA form factor 16 samples
Not Assigned area=1 circular source random sampling uniform in r and theta dA-to-dA form factor 16 samples