An asymptotically efficient closed‐form estimator for the Dirichlet distribution

Abstract

Maximum likelihood estimator (MLE) of the Dirichlet distribution is usually obtained by using the Newton–Raphson algorithm. However, in some cases, the computational costs can be burdensome, for example, in real‐time processes. Therefore, it is beneficial to develop a closed‐form estimator that is as efficient as the MLE for large sample. Here, we suggest asymptotically efficient closed‐form estimator based on the classical large sample theory.