The Bayes factor is a ratio of two competing statistical models represented by their evidence, and is used to quantify the support for one model over the other.[1] The models in question can have a common set of parameters, such as a null hypothesis and an alternative, but this is not necessary; for instance, it could also be a non-linear model compared to its linear approximation. The Bayes factor can be thought of as a Bayesian analog to the likelihood-ratio test, although it uses the integrated (i.e., marginal) likelihood rather than the maximized likelihood. As such, both quantities only coincide under simple hypotheses (e.g., two specific parameter values).[2] Also, in contrast with null hypothesis significance testing, Bayes factors support evaluation of evidence in favor of a null hypothesis, rather than only allowing the null to be rejected or not rejected.[3]
Although conceptually simple, the computation of the Bayes factor can be challenging depending on the complexity of the model and the hypotheses.[4] Since closed-form expressions of the marginal likelihood are generally not available, numerical approximations based on MCMC samples have been suggested.[5] For certain special cases, simplified algebraic expressions can be derived; for instance, the Savage–Dickey density ratio in the case of a precise (equality constrained) hypothesis against an unrestricted alternative.[6] [7] Another approximation, derived by applying Laplace's approximation to the integrated likelihoods, is known as the Bayesian information criterion (BIC);[8] in large data sets the Bayes factor will approach the BIC as the influence of the priors wanes. In small data sets, priors generally matter and must not be improper since the Bayes factor will be undefined if either of the two integrals in its ratio is not finite.
The Bayes factor is the ratio of two marginal likelihoods; that is, the likelihoods of two statistical models integrated over the prior probabilities of their parameters.[9]
\Pr(M|D)
\Pr(M|D)=
| \Pr(D|M)\Pr(M) | |
| \Pr(D) |
.
The key data-dependent term
\Pr(D|M)
Given a model selection problem in which one wishes to choose between two models on the basis of observed data D, the plausibility of the two different models M1 and M2, parametrised by model parameter vectors
\theta1
\theta2
K=
| \Pr(D|M1) | |
| \Pr(D|M2) |
=
| \int\Pr(\theta1|M1)\Pr(D|\theta1,M1)d\theta1 | |
| \int\Pr(\theta2|M2)\Pr(D|\theta2,M2)d\theta2 |
=
| ||||
|
=
| \Pr(M1|D) | |
| \Pr(M2|D) |
| \Pr(M2) | |
| \Pr(M1) |
.
When the two models have equal prior probability, so that
\Pr(M1)=\Pr(M2)
Other approaches are:
A value of K > 1 means that M1 is more strongly supported by the data under consideration than M2. Note that classical hypothesis testing gives one hypothesis (or model) preferred status (the 'null hypothesis'), and only considers evidence against it. The fact that a Bayes factor can produce evidence for and not just against a null hypothesis is one of the key advantages of this analysis method.[13]
Harold Jeffreys gave a scale for interpretation of K:[14]
! K !! dHart !! bits !! Strength of evidence|-| < 100 || < 1 || < 0 || Negative (supports M2)|-| 100 to 101/2 || 1 to 3.2 || 0 to 1.6 || Barely worth mentioning|-| 101/2 to 101 || 3.2 to 10 || 1.6 to 3.3 || Substantial|-| 101 to 103/2 || 10 to 31.6 || 3.3 to 5.0 || Strong|-| 103/2 to 102 || 31.6 to 100 || 5.0 to 6.6 || Very strong|-| > 102 || > 100 || > 6.6 || Decisive|-|}
The second column gives the corresponding weights of evidence in decihartleys (also known as decibans); bits are added in the third column for clarity. According to I. J. Good a change in a weight of evidence of 1 deciban or 1/3 of a bit (i.e. a change in an odds ratio from evens to about 5:4) is about as finely as humans can reasonably perceive their degree of belief in a hypothesis in everyday use.[15]
An alternative table, widely cited, is provided by Kass and Raftery (1995):[10]
! log10 K !! K !! Strength of evidence|-| 0 to 1/2 || 1 to 3.2 || Not worth more than a bare mention|-| 1/2 to 1 || 3.2 to 10 || Substantial|-| 1 to 2 || 10 to 100 || Strong|-| > 2 || > 100 || Decisive|-|}
Suppose we have a random variable that produces either a success or a failure. We want to compare a model M1 where the probability of success is q =, and another model M2 where q is unknown and we take a prior distribution for q that is uniform on [0,1]. We take a sample of 200, and find 115 successes and 85 failures. The likelihood can be calculated according to the binomial distribution:
{{200\choose115}q115(1-q)85
Thus we have for M1
P(X=115\midM1)={200\choose115}\left({1\over2}\right)200 ≈ 0.006
whereas for M2 we have
P(X=115\midM2)=
| 1{200 | |
| \int | |
| 0 |
\choose115}q115(1-q)85dq={1\over201} ≈ 0.005
The ratio is then 1.2, which is "barely worth mentioning" even if it points very slightly towards M1.
A frequentist hypothesis test of M1 (here considered as a null hypothesis) would have produced a very different result. Such a test says that M1 should be rejected at the 5% significance level, since the probability of getting 115 or more successes from a sample of 200 if q = is 0.02, and as a two-tailed test of getting a figure as extreme as or more extreme than 115 is 0.04. Note that 115 is more than two standard deviations away from 100. Thus, whereas a frequentist hypothesis test would yield significant results at the 5% significance level, the Bayes factor hardly considers this to be an extreme result. Note, however, that a non-uniform prior (for example one that reflects the fact that you expect the number of success and failures to be of the same order of magnitude) could result in a Bayes factor that is more in agreement with the frequentist hypothesis test.
A classical likelihood-ratio test would have found the maximum likelihood estimate for q, namely
\hatq=
| 115 | |
| 200 |
=0.575
styleP(X=115\midM2)={{200\choose115}\hatq115(1-\hatq)85
M2 is a more complex model than M1 because it has a free parameter which allows it to model the data more closely. The ability of Bayes factors to take this into account is a reason why Bayesian inference has been put forward as a theoretical justification for and generalisation of Occam's razor, reducing Type I errors.[16]
On the other hand, the modern method of relative likelihood takes into account the number of free parameters in the models, unlike the classical likelihood ratio. The relative likelihood method could be applied as follows. Model M1 has 0 parameters, and so its Akaike information criterion (AIC) value is
2 ⋅ 0-2 ⋅ ln(0.005956) ≈ 10.2467
2 ⋅ 1-2 ⋅ ln(0.056991) ≈ 7.7297
| \exp\left( | 7.7297-10.2467 |
| 2 |
\right) ≈ 0.284
. Jeff Gill (academic) . Bayesian Methods : A Social and Behavioral Sciences Approach . Chapman & Hall . 2002 . 1-58488-288-3 . Bayesian Hypothesis Testing and the Bayes Factor . 199–237 .