---
tag: YDSNHKM3EG
kind: concept
name: Frequentist inference
msc: [62A01]
specializes: [JZ2S01BHJJ]
sources:
  - url: https://en.wikipedia.org/wiki/Frequentist_inference
same_as:
  wikidata: Q2158281
  wikipedia: Frequentist inference
---

## Summary

Frequentist inference is the school of [[statistical inference|JZ2S01BHJJ]] in which the unknown parameter is a fixed quantity, not a random one.
A procedure is judged by how it behaves over hypothetical repetitions of the experiment.
That is, it is judged by its [[sampling distribution|656Y8FA3RY]] under each value of the parameter.
Confidence intervals, p-values, significance tests and unbiased estimators are its main tools.
It contrasts with [[Bayesian inference|0JWRJX2RMP]], which puts a probability distribution on the parameter.

## Explanations

### L1 human

A frequentist asks: if I used this method again and again on fresh data, how often would it give the right answer?
A $95\%$ confidence interval is a method that captures the true value in $95\%$ of repetitions, whatever that true value is.
It does not say that the true value lies in today's interval with probability $95\%$; the true value is fixed, and today's interval either contains it or not.