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

Re: NIntegrate issue with symbolic parameters

• To: mathgroup at smc.vnet.net
• Subject: [mg120577] Re: NIntegrate issue with symbolic parameters
• From: Andrew Moylan <amoylan at wolfram.com>
• Date: Sat, 30 Jul 2011 05:58:35 -0400 (EDT)
• Delivered-to: l-mathgroup@mail-archive0.wolfram.com

To prevent evaluation until the parameters are numeric, you can add a Condition to the definition for Curv. For example, change

Curv[R_] :=

to

Curv[R_] /; VectorQ[R, NumberQ] :=

Don't forget to Clear[Curv] if you have already given the old definition.

Andrew Moylan
Wolfram Research



----- Original Message -----
> From: "gopher" <gophergoon at gmail.com>
> To: mathgroup at smc.vnet.net
> Sent: Friday, July 29, 2011 10:01:55 PM
> Subject: NIntegrate issue with symbolic parameters
> 
> The docs say that NIntegrate evaluates the integrand with the
> variables being symbolic, and then repeatedly evaluates the result
> numerically. This is a problem for me as my integrand calls
> Eigenvectors which results in a symbolic calculation of the
> eigenvectors. If the variables were passed numerically instead, there
> would be no problem.
> 
> In the end, I get errors about oscillation and convergence, 'slwcon'
> and 'eincr', which should never occur since the integrand is actually
> constant! (theoretically and also confirmed by Plot3D).
> 
> Is there a way out?  Since the details are a bit long, I am putting a
> link to the file below:
> 
> https://netfiles.uiuc.edu/aroy2/www/nintegrate-issue.nb
> 
> Thanks for any help,
> Abhishek
> 
>