Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Given a distribution $F$ defined as a linear combination of Gaussian distributions:

$F = \sum_{i=1}^n C_i*N(\mu_i,\sigma_i)$ with $\sum_{i=1}^n C_i = 1$

I want to find a Gaussian function $Q = a*e^{\frac{(x-b)^2}{2c^2}}$ such that $Q(x) \geq F(x)$ $\; \forall x \in \mathbb{R}$, and $a$ is minimal.

Some context: I will be using $Q$ as envelope function/proposal distribution ($\frac{Q}{a}$) in the rejection sampling method, to generate samples from $F$. I do realize it is easy to directly draw samples from $F$, but I wish to analyze/compare different monte carlo methods on this simple example. I was wondering whether a way to determine an optimal (or good) $Q$ existed in general for this $F$. In principle I only require a few example cases, so if for some particular linear combination of Gaussian distributions (e.g. $n = 2$ or $C_i = C_j$ $\; \forall i,j$), determining the optimal $Q$ is straightforward that would be of interest too.

Thank you for your help

share|improve this question
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.