quantity

bayes factor

A statistical measure used to compare the evidence for one hypothesis against another.
13 related articles · 1 formulas

Description

The Bayes factor is a statistical quantity that quantifies the relative evidence for two competing hypotheses, typically a null hypothesis and an alternative. It is the ratio of the marginal likelihoods of the observed data under each hypothesis. A Bayes factor greater than 1 favors the alternative hypothesis, while less than 1 favors the null. It is a central tool in Bayesian inference, providing a continuous measure of evidence strength that can be updated as new data arrive.

History

The concept was introduced by Harold Jeffreys in his 1939 book 'Theory of Probability', where he also proposed the scale for interpreting its magnitude (e.g., >10 as strong evidence).

How it works

For each hypothesis, you compute the probability of the observed data given that hypothesis, averaging over all possible parameter values weighted by their prior distributions. This 'marginal likelihood' is then divided: Bayes factor = (marginal likelihood under H₁) / (marginal likelihood under H₀). Unlike p-values, the Bayes factor does not require a fixed significance threshold and can be interpreted directly as the ratio of posterior odds to prior odds. It naturally penalizes complex models, as they spread prior probability over more parameter space, reducing the average likelihood.

In practice

Bayes factors are widely used in psychology, medicine, and genetics for hypothesis testing, model selection, and meta-analysis, especially when researchers want to quantify evidence for the null hypothesis rather than merely rejecting it.

Fun fact

A Bayes factor of 20 means the data are 20 times more likely under the alternative hypothesis than under the null, but it does not mean the alternative is 20 times more probable—that also depends on the prior odds.

Concept graph →

Formulas

$$B = \prod_{k=1}^N \frac{P(D_k|H_1)}{P(D_k|H_0)}$$
Bayes factor accumulates evidence for hypotheses.
What is in the formula
BvariableBayes factor — bayes factor
D_kvariablek-th data point — data
H_1variablealternative hypothesis — hypothesis
H_0variablenull hypothesis — hypothesis
Nvariablenumber of data points — sample size
∏operatormultiplication of terms for k=1 to N — product operator

Related articles · 13

q-bio.PE2026-08-152 minarXiv:2608.14965express
A Neighbor's Calm as a Safety Signal
A lone fish sees a shadow overhead. Should it flee? If it's just a leaf, it wastes energy and gives away its position to a predator. If it's a heron, hesitation is fatal. That's the eternal trade-off of the solitary individual: quick reactions mean many false alarms, while caution means risking being eaten. In a school, it's different.
High Energy2026-08-051 minarXiv:2608.04967express
Neutron Stars to Reveal Their True Size
A cosmic "caliper" will soon measure a neutron star to within a hundred meters. For an object a couple dozen kilometers across, that's like measuring the thickness of a human hair from a kilometer away. A neutron star is the remnant of an exploded supernova.
q-bio.CB2026-08-011 minarXiv:2608.00846express
How a Cell Chooses Its Path: A View Through the Noise
A Dictyostelium amoeba crawls across glass. Its path is recorded — a chain of points on a plane. But what forces are pushing it? Inside the cell, a molecular motor works: proteins pull, the membrane springs, water presses.
Sentinel with a Map: Hunting Free-Floating Planets
Two hundred million stars. In the heart of the Galaxy, where the future survey by the Nancy Grace Roman telescope will aim, about 250 times during its mission the light of one of them will flicker at an invisible touch.
How Confirmation Bias Helps You Think Faster
We're used to thinking that noticing only confirmations of our guesses is a mistake. But recent calculations, inspired by wave behavior, flip that view. It turns out that the habit of seeking the familiar is the best way to reduce information noise (scientists call it entropy).
General Relativity2026-05-261 minarXiv:2605.27500express
Gravitational Waves Leave an Eternal Imprint
Gravitational waves are the very trembling of space, born when black holes or other massive bodies collide. Their main oddity isn't that they pass through us, but that they permanently alter the curvature of spacetime and the distances between everything they touch.
Quantum Physics2026-05-161 minarXiv:2605.16978express
Getting the Most Out of a Weak Signal
Dip your finger into water once, and from that single sensation, try to guess its temperature. In the quantum world, such one-shot measurements happen all the time, except you have to guess invisible parameters: say, the squeezing strength of a laser beam.
Instrumentation2026-05-031 minarXiv:2605.02969express
Diagnosis for Distant Worlds: How Not to Mistake Life for Noise
The hunt for life on distant planets is like diagnosing a rare disease with a test of unknown accuracy. A single positive result means nothing: maybe it's a glitch? We search the atmospheres of exoplanets (worlds around other stars) for traces of water and other compounds, but any signal could be a false alarm.
General Relativity2026-04-292 minarXiv:2604.27185express
Gravitational Waves See No Split in Black Holes
The trembling of space sounds like an orchestra. Gravitational waves caught by the LIGO and Virgo detectors tell of black hole collisions. Sometimes the music carries off-key notes — tiny deviations from Einstein's score. They might whisper of new physics or simply of dust on the bows.
Statistical Mech2026-04-251 minarXiv:2604.23346express
Measurement changes the rules of the game
Tiny magnets on a lattice, the classic Ising model, are usually described by how neighbors influence each other. But what happens when the system is closely watched? Each measurement of a coupling energy is like a gaze that forces the magnets to 'decide' their state.
Quantum Physics2026-03-171 minarXiv:2603.17030express
Comparing Scents: A New Type of Bell Inequalities
Comparing two scents is easy even without knowing their ingredients, just tell if they're alike. This everyday principle became the foundation for a new class of Bell inequalities, reimagining tests of the spooky connection between particles. Previously, you had to measure precise properties, like spin direction, but the authors showed that simply comparing outcomes is enough.
q-bio.GN2026-01-142 minarXiv:2601.09758express
How to Tell a Technical Glitch from a Real Disease
Genetic diagnostics is like comparing two photos taken with different cameras. If one camera slightly shifts the colors, the image will show "defects" that don't exist in reality.
High Energy2026-01-071 minarXiv:2601.04294express
The Secret of Neutron Stars: Sizing Them Up
Inside a neutron star, matter is compressed to unimaginable densities. Previously, scientists speculated about its exact makeup, cycling through possible equations linking pressure and density. The new method tosses out guesswork: it directly ties the measurable mass and radius to the internal structure.