We currently use the "exact" method given in Naus 1982 (which is really from Bateman 1948) -- but this method, as the paper explains, becomes "tedious" in certain cases. (Turns out that in practice "tedious" means "Python will vomit up a few runtime warnings and then return NaN.")
Naus 1982 gives some approximations which would probably help get around this problem.
We currently use the "exact" method given in Naus 1982 (which is really from Bateman 1948) -- but this method, as the paper explains, becomes "tedious" in certain cases. (Turns out that in practice "tedious" means "Python will vomit up a few runtime warnings and then return NaN.")
Naus 1982 gives some approximations which would probably help get around this problem.