Q3.36

[ytbwd]

I have confused about the example illustrating the usage of importance sampling in Monte Carlo method. For the digital option case, the Radon-Nikodym derivative has the density function f(x) = alpha/a, 0<x<a and (1-alpha)/(1-a) for a<x<1, whose expectation is 0.5+(a-alpha)/2 and will never be 1 for a belongs to (0,1). Could any one point which part is wrong? thanks!