I implore the good people to quickly glance through this thread over at StackOverflow to get a better idea of my question if the following isn't clear.
I have these integrals to evaluate using Importance Sampling--
1) x^(-0.5)
; x in [0.01,1]
2) [1+sinh(2x)ln(x)]^-1
; x in [0.8,3]
Question 1 How does one chose the importance PDF really? For the above functions I just plotted exponential and normal functions (since it just occurred to me that they are similar to the integrals to be evaluated. Otherwise, is there a procedure to arrive at which PDF is to be used?
Question 2 Is it necessary that the chosen importance PDF has to have lesser variance than the one to be evaluated? My understanding is that the PDF chosen should follow and cover the plot of the given integral.
Question 3
So, I have a function x^2 + y^2
, where x,y in[-1,1]
. For cases like this, where the integral to be evaluated is a double integral, what PDF should i think about? 2-D normal?
If it's a random function, there's no way to generate samples from it (for the lack of built in commands in R- only standard PDFs exist.)
self-study
tag, and try to follow the suggested guidelines there. (You'd need to remove one of your 5 current tags first though. I think two of them aren't really needed.) $\endgroup$