The problem is that there is no explicit definition for the values you are looking for.
In my Balanis (3rd Edition), page 821, a rule of thumb is given that you h/lambda0 < 0.1. The largest impact of the dielectric height is the effective dielectric constant, for which he also gives equation 14-1.
To my knowledge, the ground plane size really just determines the amount of 'ripple' seen in the radiation pattern's main lobe. The bigger, the better, but of course practicality has to factor in somewhere. From my (limited) experience, 6x(L,W) should be fairly good. People with more expertise on patches, of which there is at least one on these forums, may be able to give you a few more pointers.