\(F^{-1}(u)\) can be obtained by integrating \(f(x)\), setting the result equal to \(u\), and solving for \(x\):
$$\begin{array}{c} f(x) = e{-x2/2^2}
F(x) = f(x) dx = - e{-x2/2^2} = u
F^{-1}(u) = \end{array}$$
Formula
# For given sigma the value can be estimated using antithetic Variables.
# When getting m samples, X1 represents value generated from m/2 samples from uniform distribution u1
# Similarly X-dash from anithetic sample u-dash = 1 - u
# X2 represents value generated from m/2 samples of uniform distribution u2
anti = function(m,sigma){
u1 = runif(m/2)
x1 = sigma * sqrt(-2 * log(u1))
udash = 1 - u1
xdash = sigma * sqrt(-2 * log(udash))
#second samples
u2 = runif(m/2)
x2 = sigma * sqrt(-2 * log(u2))
#Applying the variance formula
vardash = (var(x1) + var(xdash) + 2 * cov(x1, xdash)/4)
var2 = (var(x1) + var(x2) + 2 * cov(x1, x2)/4)
#find the % reduction in varianace
return((var2 - vardash)/var2)
}
anti(1000,1)
## [1] 0.2021825
Reference : Referred other students discussion.