Bayes factor
E700155
The Bayes factor is a Bayesian model comparison metric that quantifies how much more strongly data support one statistical model or hypothesis over another.
All labels observed (1)
| Label | Occurrences |
|---|---|
| Bayes factor canonical | 1 |
Statements (48)
| Predicate | Object |
|---|---|
| instanceOf |
Bayesian model comparison metric
ⓘ
likelihood ratio ⓘ statistical measure ⓘ |
| alternativeName | Bayesian evidence ratio NERFINISHED ⓘ |
| appliedIn |
biostatistics
ⓘ
econometrics ⓘ machine learning ⓘ physics ⓘ psychology ⓘ |
| approximatedBy | BIC difference ⓘ |
| compares |
two hypotheses
ⓘ
two statistical models ⓘ |
| computedBy |
Laplace approximation
NERFINISHED
ⓘ
Monte Carlo methods ⓘ bridge sampling ⓘ reversible jump MCMC ⓘ |
| contrastsWith |
frequentist hypothesis tests
ⓘ
p-values ⓘ |
| definedAs | ratio of marginal likelihoods of two models ⓘ |
| dependsOn |
likelihood function
ⓘ
prior distributions ⓘ |
| domain | statistical inference ⓘ |
| hasFormula | B_10 = p(y | M_1) / p(y | M_0) ⓘ |
| hasProperty |
can accumulate evidence for the null model
ⓘ
can be computed analytically in simple models ⓘ often requires numerical integration ⓘ penalizes model complexity via marginalization ⓘ sensitive to prior specification ⓘ |
| interpretedAs | how much data update prior odds to posterior odds ⓘ |
| interpretedUsing |
Jeffreys scale
NERFINISHED
ⓘ
Kass and Raftery scale NERFINISHED ⓘ |
| introducedBy | Harold Jeffreys NERFINISHED ⓘ |
| invariantUnder | reparameterization of model parameters ⓘ |
| measures | strength of evidence in data for one model over another ⓘ |
| quantifies | relative evidence from data ⓘ |
| relatedTo |
Bayesian Information Criterion
NERFINISHED
ⓘ
marginal likelihood ⓘ posterior odds ⓘ prior odds ⓘ |
| symbol |
B_10
ⓘ
K ⓘ |
| usedFor |
hypothesis testing
ⓘ
model comparison ⓘ |
| usedIn | Bayesian statistics NERFINISHED ⓘ |
| usedTo |
assess evidence for alternative hypothesis
ⓘ
assess evidence for null hypothesis ⓘ select between nested models ⓘ select between non-nested models ⓘ |
Referenced by (1)
Full triples — surface form annotated when it differs from this entity's canonical label.