frequentiste

Frequentist statistics, sometimes referred to in French as frequentiste, is a school of statistical inference that interprets probability as the long-run frequency of outcomes in repeated samples from a population. In this view, parameters are fixed but unknown, and uncertainty about them is quantified through the behavior of data under repeated experimentation rather than by personal belief.

Key tools include hypothesis testing, confidence intervals, and p-values. Hypothesis tests assess whether observed data are unlikely under a null hypothesis. Confidence intervals are constructed to have a specified coverage probability in the long run. A p-value measures the probability, under the null, of obtaining data as extreme as observed.

Historically, frequentist ideas were developed by figures such as Ronald Fisher, Jerzy Neyman, and Egon Pearson, who emphasized procedures with controlled long-run error rates and reliance on sampling distributions.

Frequentists distinguish probability statements about data from beliefs about parameters. They typically treat parameters as fixed, whereas Bayesian methods treat them as random variables with prior distributions.

Critics argue that p-values can be misleading and that confidence intervals are often misinterpreted. Challenges include multiple testing, model misspecification, and dependence among data, which can distort error rates and inference in practice.

Frequentist methods remain standard in many scientific fields, especially where objective error control under repeated sampling is valued. They underpin common statistical procedures in textbooks and software.