[Date Index] [Thread Index] [Author Index]

integration for CDF[NoncentralChiSquareDistribution] failing

• To: mathgroup at smc.vnet.net
• Subject: [mg74806] integration for CDF[NoncentralChiSquareDistribution] failing
• From: Albert Maydeu-Olivares <amaydeu at ub.edu>
• Date: Fri, 6 Apr 2007 04:17:39 -0400 (EDT)

Hi, 

The built-in function appears to work well for small number of degrees of freedom. For large df, integration converges too slowly and often fails. For instance

1 - CDF[NoncentralChiSquareDistribution[57, 5], 1]

yields 1.04, when as a probability it should be < 1. Does anybody have a function that works better than the built-in? Also, has anybody implemented a normal approximation? I had planned to work with 300+ df. 

Albert

• Follow-Ups:
  • Re: integration for CDF[NoncentralChiSquareDistribution]
    • From: Darren Glosemeyer <darreng@wolfram.com>