In summary, In order to find the average value of ##\varepsilon(\alpha_1,\,\alpha_2,\,\alpha_3)##, you use the cumulative distribution function (CDF) of X, the derivative with respect to x, and the probability density function (PDF).
#36
EngWiPy
1,368
61
Is it an error margin in the numerical evaluation of the integral?
Upload image to a hosting site (dropbox, google photos, tumblr, flickr, etc) then click to "image" on the toolbar and insert the image url.
#40
EngWiPy
1,368
61
It doesn't work. I tried both dropbox and google photos. Isn't it "get a link" that I need to insert here?
#41
EngWiPy
1,368
61
Attached is the figure. Sim=Monte-Carlo simulations
Attachments
untitled.jpg
24.6 KB
· Views: 430
#42
EngWiPy
1,368
61
I think Monte-Carlo simulations are accurate, so, I suspect, given that everything else done properly, the accuracy of NIntegrate in Mathematica is the issue. Is there any possible reason and I cannot see it?
#43
Hawkeye18
178
62
Monte-Carlo convergence is quite slow, ##C/\sqrt N##, maybe the error is due to that. Also there could be some details that Monte-Carlo misses.
On the other hand how Mathematica does NIntegrate is hidden, it might also introduce some systematic error here.
#44
EngWiPy
1,368
61
I remember using Mathematica to evaluate some numerical integral in my master thesis. There was no difference between the numerical integration and Monte-Carlo simulations then!