Response to the following problem:
(O Ruanaidh and Fitzgerald, 1996) For simulating random variables ´ from the density
F(X) ∝ exp {-x2 √x}[sin(x)2] , 0
compare the following choices of instrumental densities on R:
g1(x) = (1/2)e-|x| , g2(x) = (1/2π)(1/1+x2/4) , g3(x) = (1/√2π)e-x2/2
For each of the instrumental densities, estimate the number M of simulations needed to obtain three digits of accuracy in estimating Ef [X].
Deduce from the acceptance rate an estimator of the normalizing constant of f for each of the instrumental densities.