#4 Noncentral f distribution defective

[LouTheBlueGuru]

The noncentral cumulative f distribution only works for select parameters. It appears to be most sensitive to the non centrality parameter. At some point it returns probabilities > 1 and for all greater values of the non centrality parameter returns nonsense from this point on! I have traced the problem back to the noncentral cumulative beta distribution (and maybe even further back to the gamma distribution). There is a piece of suspicious code that uses a user defined C fmax2 function that was replaced with the Java Math.max function, but since I do not have the original C code, I have no idea what is really going on. This collection of stat routines would be very useful if they actually worked!