4
$\begingroup$

Usually Monte Carlo method is used to compute integration. For example, let $g(x,\theta)$ be a continuous function about $x$ and $\theta$, $f(x \mid \theta)$ is a continuous pdf with parameter $\theta$. if the integral $E[g(x,\theta)]=\int g(x,\theta)f(x\mid \theta)dx$ has no closed form . Then we can use Monte Carlo approximation $E[g(x,\theta)] \approx \frac{1}{m}\sum_{l=1}^m g(X_i,\theta)$, where $X_1,\dots,X_m$ is a random sample from distribution $f(x \mid \theta)$. By law of large number, $\frac{1}{m}\sum_{l=1}^m g(X_i,\theta)$ converges to $E[g(x,\theta)]$ in probability for each $\theta \in \Theta$. Do we have theorem talking about uniform convergence of Monte Carlo approximation? I mean, under what condition, we have

$\sup\limits_{\theta\in\Theta} \left\| \frac1n\sum_{i=1}^n g(X_i,\theta) - \operatorname{E}[g(X,\theta)] \right\| \xrightarrow \ 0$ a.s. or in probability

Assume parameter space $\Theta$ is compact

$\endgroup$

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.