Skip to content
Get started

The best Laplace envelope for a normal

You need standard normal samples but can only cheaply draw from a Laplace (double-exponential) law with density gλ(x)=λ2e−λ∣x∣g_\lambda(x)=\tfrac{\lambda}{2}e^{-\lambda|x|}. You use rejection sampling: draw X∼gλX\sim g_\lambda and accept it with probability φ(X)/(M gλ(X))\varphi(X)/\big(M\,g_\lambda(X)\big), where φ\varphi is the standard normal density and MM is the smallest constant with φ≤Mgλ\varphi\le M g_\lambda everywhere. You may choose the rate λ>0\lambda>0 freely. What is the highest achievable long-run acceptance rate? Give a decimal to four places.

Show hints (2)+
  1. Show that the overall acceptance probability is 1/M1/M, whatever the proposal density.
  2. Find M(λ)=sup⁡xφ(x)/gλ(x)M(\lambda)=\sup_x \varphi(x)/g_\lambda(x) (the sup sits at ∣x∣=λ|x|=\lambda), then minimise over λ\lambda.

Answer

Reveal answer →

0.7602 (± 0.0005)

Create a free account to see the full worked solution and track your progress.

Interview format: Prop Trading, Systematic & Data-Driven Trading

Related questions