Re: please explain why numerical integration is attempted

*To*: mathgroup at smc.vnet.net*Subject*: [mg72531] Re: [mg72492] please explain why numerical integration is attempted*From*: Darren Glosemeyer <darreng at wolfram.com>*Date*: Fri, 5 Jan 2007 01:58:08 -0500 (EST)*References*: <200701030542.AAA29349@smc.vnet.net> <459D1B60.5080501@wolfram.com> <acbec1a40701040755l5d0a36dbud0a3d3545ce3723@mail.gmail.com>

Yes, unevaluated would be the best result in this case. N can then be used to get a numeric result if desired. This would be consistent with, for instance, special functions that are defined by an integral that cannot be written in some other nice form for certain exact numeric values (e.g. Beta[Pi,E]). Darren Glosemeyer Wolfram Research Chris Chiasson wrote: > Thanks for the reply. > > Why doesn't it just return unevaluated or with a symbolic integral? > > On 1/4/07, Darren Glosemeyer <darreng at wolfram.com> wrote: >> Chris Chiasson wrote: >> > Needs["Statistics`MultinormalDistribution`"] >> > CDF[MultinormalDistribution[{1, >> > >> 3},{{1/50,-2/125},{-2/125,9/500}},{{1/(5*Sqrt[2]),0},{(-2*Sqrt[2])/25, >> > Sqrt[13]/50}}],{2/10,1}] >> > >> > Thank you, >> > >> > >> Closed form results are generally not known for cdfs of multivariate >> normal distributions with non-diagonal covariance matrices. Numeric >> methods are needed to obtain a result. >> >> Darren Glosemeyer >> Wolfram Research >> > >

**References**:**please explain why numerical integration is attempted for this input***From:*"Chris Chiasson" <chris@chiasson.name>