Maximum Likelihood Estimator

What is the maximum likelihood estimator (MLE) and how does it work?

Answer

Suppose we have $x_1, \dots, x_n$ i.i.d. observations from a distribution with density (or pmf) $f(x;\theta)$. The maximum likelihood estimator is the value of $\hat{\theta}$ for which the probability of observing these exact data points again is maximized. This is the same as finding the $\theta$ that maximizes the likelihood function. Formally, the MLE is defined as

\[ \hat{\theta}_{\text{MLE}} = \arg\max_{\theta} L(\theta). \]

In practice, we usually maximize the log-likelihood

\[ \ell(\theta) = \log L(\theta) = \sum_{i=1}^n \log f(x_i;\theta), \]

since the log turns products into sums (making derivatives easier) and is monotone (preserving the order/maximizer).

Back to collection