Gene clustering: 2001: 2
Number of Trials
§User sets desired confidence Þ ptarg 
§Need enough trials to be sure that
pest > ptarg   or   pest< ptarg 
•More trials Þ better estimate
•Bernoulli experiment  Þ estimate standard error
•Repeat until estimate is confidently below/above target
§Choosing candidate minimum size
•Do binary search on possible minimum sizes
•Compute pest for every size in single run.