Re: Strange error message from NIntegrate
- To: mathgroup at smc.vnet.net
- Subject: [mg28159] Re: [mg28149] Strange error message from NIntegrate
- From: BobHanlon at aol.com
- Date: Mon, 2 Apr 2001 02:18:29 -0400 (EDT)
- Sender: owner-wri-mathgroup at wolfram.com
You made the power of x a list (brackets vice parenthreses).
There are still problems with the numerical integration. However, you can
- approximate the answer with a finite upper limit of integration
- change the number of intervals (GaussPoints)
- change the method of numerical integration
NIntegrate[x^(3/2)*Exp[-x]*BesselI[3/2,x/2], {x,0, 1000}]
1.0030037040845279
NIntegrate[x^(3/2)*Exp[-x]*BesselI[3/2,x/2], {x,0, Infinity},
GaussPoints -> 10]
1.0030037040849122
NIntegrate[x^(3/2)*Exp[-x]*BesselI[3/2,x/2], {x,0, Infinity},
Method -> Trapezoidal]
1.0030037040847963
Bob Hanlon
In a message dated 2001/4/1 1:16:26 AM, h.holm at spray.no writes:
>I am trying to do the following numerical integration:
>
> NIntegrate[x^{3/2}*Exp[-x]*BesselI[3/2,x/2], {x,0, Infinity}]
>
>this results in the following error,
>
> NIntegrate::"inum":
> Integrand {0.141859424324175} is not numerical at {x} = {1.}.
>
>I think this is very strange, after all, 0.1418... _is_ a
>number. It might be noted that doing the integration
>symbolically works nicely:
>
> Integrate[x^{3/2}*Exp[-x]*BesselI[3/2,x/2], {x,0, Infinity}]
>
>which gives the following result:
>
> 16/(9*Sqrt[pi])
>
>
>I guess I am doing something wrong, but I cannot understand
>what. Can anybody help me?
>