Brown Bag Session | Monte Carlo Analysis and Bootstrapping in Stata
Dr. Ramon Christen
The seminar aims to familiarize participants with two concepts in statistics that use simulation techniques. Foreknowledge in linear regression and Stata are a plus but not necessary.
When properties of an estimator cannot be derived analytically, we can use simulation methods. Monte Carlo analyses draw pseudo-samples from an pre-defined stochastic data generating process and analyze the behavior of an estimator when parameters (e.g. sample size) are modified. This enables to study the accuracy of an estimator when the sample size is small, for instance.
Bootstrapping uses a similar idea. It provides estimates of a sample distribution when analytical solutions are not feasible. A typical field of application is the estimation confidence intervals for unknown distributions.