Hipnotic Bayestas

Bayesian Inference
Statisticat, LLC
Abstract
The Bayesian interpretation of probability is one of two broad categories of interpre-
tations. Bayesian inference updates knowledge about unknowns, parameters, with infor-
mation from data. The LaplacesDemon package is a complete environment for Bayesian
inference within R, and this vignette provides an introduction to the topic. This arti-
cle introduces Bayes’ theorem, model-based Bayesian inference, components of Bayesian
inference, prior distributions, hierarchical Bayes, conjugacy, likelihood, numerical approx-
imation, prediction, Bayes factors, model fit, posterior predictive checks, and ends by
comparing advantages and disadvantages of Bayesian inference.
Keywords: Bayesian, LaplacesDemon, LaplacesDemonCpp, R.
This article is an introduction to Bayesian inference for users of the LaplacesDemon package
(Statisticat LLC. 2015) in R (R Core Team 2014), often referred to as LD. LaplacesDemonCpp
is an extension package that uses C++. A formal introduction to LaplacesDemon is provided
in an accompanying vignette entitled “LaplacesDemon Tutorial”. Merriam-Webster defines
‘Bayesian’ as follows
Bayesian : being, relating to, or involving statistical methods that assign proba-
bilities or distributions to events (as rain tomorrow) or parameters (as a population
mean) based on experience or best guesses before experimentation and data col-
lection and that apply Bayes’ theorem to revise the probabilities and distributions
after obtaining experimental data.
In statistical inference, there are two broad categories of interpretations of probability: Bayesian
inference and frequentist inference. These views often differ with each other on the fundamen-
tal nature of probability. Frequentist inference loosely defines probability as the limit of an
event’s relative frequency in a large number of trials, and only in the context of experiments
that are random and well-defined. Bayesian inference, on the other hand, is able to assign
probabilities to any statement, even when a random process is not involved. In Bayesian
inference, probability is a way to represent an individual’s degree of belief in a statement, or
given evidence.
Within Bayesian inference, there are also different interpretations of probability, and different
approaches based on those interpretations. The most popular interpretations and approaches
2 Bayesian Inference
are objective Bayesian inference (Berger 2006) and subjective Bayesian inference (Anscombe
and Aumann 1963; Goldstein 2006). Objective Bayesian inference is often associated with
Bayes and Price (1763), Laplace (1814), and Jeffreys (1961). Subjective Bayesian inference is
often associated with Ramsey (1926), De Finetti (1931), and Savage (1954). The first major
event to bring about the rebirth of Bayesian inference was De Finetti (1937). Differences in
the interpretation of probability are best explored outside of this article1 .
This article is intended as an approachable introduction to Bayesian inference, or as a handy
summary for experienced Bayesians. It is assumed that the reader has at least an elemen-
tary understanding of statistics, and this article focuses on applied, rather than theoretical,
material. Equations and statistical notation are included, but it is hopefully presented so the
reader does not need an intricate understanding of solving integrals, for example, but should
understand the basic concept of integration. Please be aware that it is difficult to summarize
Bayesian inference in such a short article. In which case, consider Robert (2007) for a more
thorough and formal introduction.
1. Bayes’ Theorem
Bayes’ theorem shows the relation between two conditional probabilities that are the reverse
of each other. This theorem is named after Reverend Thomas Bayes (1701-1761), and is also
referred to as Bayes’ law or Bayes’ rule (Bayes and Price 1763)2 . Bayes’ theorem expresses
the conditional probability, or ‘posterior probability’, of an event A after B is observed in
terms of the ‘prior probability’ of A, prior probability of B, and the conditional probability
of B given A. Bayes’ theorem is valid in all common interpretations of probability. The two
(related) examples below should be sufficient to introduce Bayes’ theorem.
1.1. Bayes’ Theorem, Example 1

Bayes’ theorem provides an expression for the conditional probability of A given B, which is
equal to
Pr(B|A) Pr(A)
Pr(A|B) = (1)
Pr(B)
For example, suppose one asks the question: what is the probability of going to Hell, condi-
tional on consorting (or given that a person consorts) with Laplace’s Demon3 . By replacing
A with Hell and B with Consort, the question becomes
Pr(Consort|Hell) Pr(Hell)
Pr(Hell|Consort) =
Pr(Consort)
1
If these terms are new to the reader, then please do not focus too much on the words ‘objective’ and
‘subjective’, since there is a lot of debate over them. For what it’s worth, Statisticat, LLC, the provider of this
R package entitled LaplacesDemon, favors the ‘subjective’ interpretation.
2
Stigler (1983) suggests the earliest discoverer of Bayes’ theorem was Nicholas Saunderson (1682-1739), a
blind mathematician/optician, who at age 29 became the Lucasian Professor of Mathematics at Cambridge.
This position was previously held by Isaac Newton.
3
This example is, of course, intended with humor.
Statisticat LLC 3
Note that a common fallacy is to assume that Pr(A|B) = Pr(B|A), which is called the
conditional probability fallacy.
1.2. Bayes’ Theorem, Example 2

Another way to state Bayes’ theorem is
Pr(B|Ai ) Pr(Ai )
Pr(Ai |B) =
Pr(B|Ai ) Pr(Ai ) + ... + Pr(B|An ) Pr(An )
Let’s examine our burning question, by replacing Ai with Hell or Heaven, and replacing B
with Consort
• Pr(A1 ) = Pr(Hell)
• Pr(A2 ) = Pr(Heaven)
• Pr(B) = Pr(Consort)
• Pr(A1 |B) = Pr(Hell|Consort)
• Pr(A2 |B) = Pr(Heaven|Consort)
• Pr(B|A1 ) = Pr(Consort|Hell)
• Pr(B|A2 ) = Pr(Consort|Heaven)
Laplace’s Demon was conjured and asked for some data. He was glad to oblige.
Data
• 6 people consorted out of 9 who went to Hell.
• 5 people consorted out of 7 who went to Heaven.
• 75% of the population goes to Hell.
• 25% of the population goes to Heaven.
Now, Bayes’ theorem is applied to the data. Four pieces are worked out as follows
• Pr(Consort|Hell) = 6/9 = 0.666
• Pr(Consort|Heaven) = 5/7 = 0.714
• Pr(Hell) = 0.75
• Pr(Heaven) = 0.25
Finally, the desired conditional probability Pr(Hell|Consort) is calculated using Bayes’ theo-
rem
0.666(0.75)
• Pr(Hell|Consort) = 0.666(0.75)+0.714(0.25)
• Pr(Hell|Consort) = 0.737
The probability of someone consorting with Laplace’s Demon and going to Hell is 73.7%,
which is less than the prevalence of 75% in the population. According to these findings,
consorting with Laplace’s Demon does not increase the probability of going to Hell. With
that in mind, please continue. . .
2. Model-Based Bayesian Inference

The basis for Bayesian inference is derived from Bayes’ theorem. Here is Bayes’ theorem,
equation 1, again
Pr(B|A) Pr(A)
Pr(A|B) =
Pr(B)
Replacing B with observations y, A with parameter set Θ, and probabilities Pr with densities
p (or sometimes π or function f ), results in the following
p(y|Θ)p(Θ)
p(Θ|y) =
p(y)
where p(y) will be discussed below, p(Θ) is the set of prior distributions of parameter set
Θ before y is observed, p(y|Θ) is the likelihood of y under a model, and p(Θ|y) is the joint
posterior distribution, sometimes called the full posterior distribution, of parameter set Θ
that expresses uncertainty about parameter set Θ after taking both the prior and data into
account. Since there are usually multiple parameters, Θ represents a set of j parameters, and
may be considered hereafter in this article as
Θ = θ1 , ..., θj
The denominator
Z
p(y) = p(y|Θ)p(Θ)dΘ
defines the “marginal likelihood” of y, or the “prior predictive distribution” of y, and may be
set to an unknown constant c. The prior predictive distribution4 indicates what y should
look like, given the model, before y has been observed. Only the set of prior probabilities and
the model’s likelihood function are used for the marginal likelihood of y. The presence of the
marginal likelihood of y normalizes the joint posterior distribution, p(Θ|y), ensuring it is a
proper distribution and integrates to one.
By replacing p(y) with c, which is short for a ‘constant of proportionality’, the model-based
formulation of Bayes’ theorem becomes
p(y|Θ)p(Θ)
p(Θ|y) =
c
4
The predictive distribution was introduced by Jeffreys (1961).
Statisticat LLC 5
By removing c from the equation, the relationship changes from ’equals’ (=) to ’proportional
to’ (∝)5
p(Θ|y) ∝ p(y|Θ)p(Θ) (2)

This form can be stated as the unnormalized joint posterior being proportional to the like-
lihood times the prior. However, the goal in model-based Bayesian inference is usually not
to summarize the unnormalized joint posterior distribution, but to summarize the marginal
distributions of the parameters. The full parameter set Θ can typically be partitioned into
Θ = {Φ, Λ}
where Φ is the sub-vector of interest, and Λ is the complementary sub-vector of Θ, often
referred to as a vector of nuisance parameters. In a Bayesian framework, the presence of
nuisance parameters does not pose any formal, theoretical problems. A nuisance parameter
is a parameter that exists in the joint posterior distribution of a model, though it is not a
parameter of interest. The marginal posterior distribution of φ, the parameter of interest,
can simply be written as
Z
p(φ|y) = p(φ, Λ|y)dΛ
In model-based Bayesian inference, Bayes’ theorem is used to estimate the unnormalized joint
posterior distribution, and finally the user can assess and make inferences from the marginal
posterior distributions.
3. Components of Bayesian Inference

The components6 of Bayesian inference are
1. p(Θ) is the set of prior distributions for parameter set Θ, and uses probability as a
means of quantifying uncertainty about Θ before taking the data into account.
2. p(y|Θ) is the likelihood or likelihood function, in which all variables are related in a full
probability model.
3. p(Θ|y) is the joint posterior distribution that expresses uncertainty about parameter
set Θ after taking both the prior and the data into account. If parameter set Θ is
partitioned into a single parameter of interest φ and the remaining parameters are
considered nuisance parameters, then p(φ|y) is the marginal posterior distribution.
4. Prior Distributions
5
For those unfamiliar with ∝, this symbol simply means that two quantities are proportional if they vary
in such a way that one is a constant multiplier of the other. This is due to the constant of proportionality c
in the equation. Here, this can be treated as ‘equal to’.
6
In Bayesian decision theory, an additional component exists (Robert 2007, p. 53), the loss function,
L(Θ, ∆).
In Bayesian inference, a prior probability distribution, often called simply the prior, of an
uncertain parameter θ or latent variable is a probability distribution that expresses uncertainty
about θ before the data are taken into account7 . The parameters of a prior distribution are
called hyperparameters, to distinguish them from the parameters (Θ) of the model.
When applying Bayes’ theorem, the prior is multiplied by the likelihood function and then
normalized to estimate the posterior probability distribution, which is the conditional distri-
bution of Θ given the data. Moreover, the prior distribution affects the posterior distribution.
Prior probability distributions have traditionally belonged to one of two categories: informa-
tive priors and uninformative priors. Here, four categories of priors are presented according
to information8 and the goal in the use of the prior. The four categories are informative,
weakly informative, least informative, and uninformative.
4.1. Informative Priors

When prior information is available about θ, it should be included in the prior distribution of
θ. For example, if the present model form is similar to a previous model form, and the present
model is intended to be an updated version based on more current data, then the posterior
distribution of θ from the previous model may be used as the prior distribution of θ for the
present model.
In this way, each version of a model is not starting from scratch, based only on the present
data, but the cumulative effects of all data, past and present, can be taken into account. To
ensure the current data do not overwhelm the prior, Ibrahim and Chen (2000) introduced the
power prior. The power prior is a class of informative prior distribution that takes previous
data and results into account. If the present data is very similar to the previous data, then the
precision of the posterior distribution increases when including more and more information
from previous models. If the present data differs considerably, then the posterior distribution
of θ may be in the tails of the prior distribution for θ, so the prior distribution contributes
less density in its tails. Hierarchical Bayes is also a popular way to combine data sets.
Sometimes informative prior information is not simply ready to be used, such as when it resides
in another person, as in an expert. In this case, their personal beliefs about the probability of
the event must be elicited into the form of a proper probability density function. This process
is called prior elicitation.
4.2. Weakly Informative Priors

Weakly Informative Prior (WIP) distributions use prior information for regularization9 and
stabilization, providing enough prior information to prevent results that contradict our knowl-
7
One so-called version of Bayesian inference is ‘empirical Bayes’, which sounds enticing because anything
‘empirical’ seems desirable. However, empirical Bayes is a term for the use of data-dependent priors, where
the prior is first modeled usually with maximum likelihood and then used in the Bayesian model. This is an
undesirable double-use of the data and is most problematic with small sample sizes (Berger 2006). It also
seems to violate the elementary concept that a prior probability distribution expresses uncertainty about θ
before the data are taken into account. It has been claimed that “empirical Bayes methods are not Bayesian”
(Bernardo 2008).
8
‘Information’ is used loosely here to describe either the prior information from personal beliefs or
informational-theoretic content.
9
The definition of regularization is to introduce additional information in order to solve an ill-posed problem
or to prevent overfitting.
Statisticat LLC 7
edge or problems such as an algorithmic failure to explore the state-space. Another goal is for
WIPs to use less prior information than is actually available. A WIP should provide some of
the benefit of prior information while avoiding some of the risk from using information that
doesn’t exist. WIPs are the most common priors in practice, and are favored by subjective
Bayesians.
Selecting a WIP can be tricky. WIP distributions should change with the sample size, because
a model should have enough prior information to learn from the data, but the prior information
must also be weak enough to learn from the data.
Following is an example of a WIP in practice. It is popular, for good reasons, to center and
scale all continuous predictors (Gelman 2008). Although centering and scaling predictors is
not discussed here, it should be obvious that the potential range of the posterior distribution
of θ for a centered and scaled predictor should be small. A popular WIP for a centered and
scaled predictor may be
θ ∼ N (0, 10000)
where θ is normally-distributed according to a mean of 0 and a variance of 10,000, which is

equivalent to a standard deviation of 100, or precision of 1.0E-4. In this case, the density for
θ is nearly flat. Nonetheless, the fact that it is not perfectly flat yields good properties for
numerical approximation algorithms. In both Bayesian and frequentist inference, it is possible
for numerical approximation algorithms to become stuck in regions of flat density, which
become more common as sample size decreases or model complexity increases. Numerical
approximation algorithms in frequentist inference function as though a flat prior were used,
so numerical approximation algorithms in frequentist inference become stuck more frequently
than numerical approximation algorithms in Bayesian inference. Prior distributions that are
not completely flat provide enough information for the numerical approximation algorithm to
continue to explore the target density, the posterior distribution.
After updating a model in which WIPs exist, the user should examine the posterior to see if the
posterior contradicts knowledge. If the posterior contradicts knowledge, then the WIP must
be revised by including information that will make the posterior consistent with knowledge
(Gelman 2008).
A popular objective Bayeisan criticism against WIPs is that there is no precise, mathematical
form to derive the optimal WIP for a given model and data.
Vague Priors
A vague prior, also called a diffuse prior10 , is difficult to define, after considering WIPs.
The first formal move from vague to weakly informative priors is Lambert, Sutton, Burton,
Abrams, and Jones (2005). After conjugate priors were introduced (Raiffa and Schlaifer
1961), most applied Bayesian modeling has used vague priors, parameterized to approximate
the concept of uninformative priors (better considered as least informative priors, see section
4.3). For more information on conjugate priors, see section 6.
Typically, a vague prior is a conjugate prior with a large scale parameter. However, vague
priors can pose problems when the sample size is small. Most problems with vague priors and
small sample size are associated with scale, rather than location, parameters. The problem
10
Some sources refer to diffuse priors as flat priors.
can be particularly acute in random-effects models, and the term random-effects is used rather
loosely here to imply exchangeable11 , hierarchical, and multilevel structures. A vague prior
is defined here as usually being a conjugate prior that is intended to approximate an uninfor-
mative prior (or actually, a least informative prior), and without the goals of regularization
and stabilization.
4.3. Least Informative Priors

The term ‘Least Informative Priors’, or LIPs, is used here to describe a class of prior in which
the goal is to minimize the amount of subjective information content, and to use a prior that
is determined solely by the model and observed data. The rationale for using LIPs is often
said to be ‘to let the data speak for themselves’. LIPs are favored by objective Bayesians.
Flat Priors
The flat prior was historically the first attempt at an uninformative prior. The unbounded,
uniform distribution, often called a flat prior, is
θ ∼ U(−∞, ∞)
where θ is uniformly-distributed from negative infinity to positive infinity. Although this

seems to allow the posterior distribution to be affected soley by the data with no impact
from prior information, this should generally be avoided because this probability distribution
is improper, meaning it will not integrate to one since the integral of the assumed p(θ) is
infinity (which violates the assumption that the probabilities sum to one). This may cause
the posterior to be improper, which invalidates the model.
Reverend Thomas Bayes (1701-1761) was the first to use inverse probability (Bayes and
Price 1763), and used a flat prior for his billiard example so that all possible values of θ
are equally likely a priori (Gelman, Carlin, Stern, and Rubin 2004, p. 34-36). Pierre-Simon
Laplace (1749-1827) also used the flat prior to estimate the proportion of female births in a
population, and for all estimation problems presented or justified as a reasonable expression
of ignorance. Laplace’s use of this prior distribution was later referred to as the ‘principle of
indifference’ or ‘principle of insufficient reason’, and is now called the flat prior (Gelman et al.
2004, p. 39). Laplace was aware that it was not truly uninformative, and used it as a LIP.
Another problem with the flat prior is that it is not invariant to transformation. For example,
a flat prior on a standard deviation parameter is not also flat for its variance or precision.
Hierarchical Prior
A hierarchical prior is a prior in which the parameters of the prior distribution are estimated
from data via hyperpriors, rather than with subjective information (Gelman 2008). Parame-
ters of hyperprior distributions are called hyperparameters. Subjective Bayesians prefer the
hierarchical prior as the LIP, and the hyperparameters are usually specified as WIPs. Hier-
archical priors are presented later in more detail in the section entitled ‘Hierarchical Bayes’.
11
For more information on exchangeability, see https://web.archive.org/web/20150418134644/http://
www.bayesian-inference.com/exchangeability.
Statisticat LLC 9
Jeffreys Prior
Jeffreys prior, also called Jeffreys rule, was introduced in an attempt to establish a least
informative prior that is invariant to transformations (Jeffreys 1961). Jeffreys prior works
well for a single parameter, but multi-parameter situations may have inappropriate aspects
accumulate across dimensions to detrimental effect.
MAXENT
A MAXENT prior, proposed by Jaynes (1968), is a prior probability distribution that is
selected among other candidate distributions as the prior of choice when it has the maximum
entropy (MAXENT) in the considered set, given constraints on the candidate set. More
entropy is associated with less information, and the least informative prior is preferred as a
MAXENT prior. The principle of minimum cross-entropy generalizes MAXENT priors from
mere selection to updating the prior given constraints while seeking the maximum, possible
entropy.
Reference Priors
Introduced by Bernardo (1979), reference priors do not express personal beliefs. Instead,
reference priors allow the data to dominate the prior and posterior (Berger, Bernardo, and
Dongchu 2009). Reference priors are estimated by maximizing the expected intrinsic discrep-
ancy between the posterior distribution and prior distribution. This maximizes the expected
posterior information about y when the prior density is p(y). In some sense, p(y) is the ‘least
informative’ prior about y (Bernardo 2005). Reference priors are often the objective prior
of choice in multivariate problems, since other rules (e.g., Jeffreys rule) may result in priors
with problematic behavior. When reference priors are used, the analysis is called reference
analysis, and the posterior is called the reference posterior.
Subjective Bayesian criticisms of reference priors are that the concepts of regularization and
stabilization are not taken into account, results that contradict knowledge are not prevented,
a numerical approximation algorithm may become stuck in low-probability or flat regions,
and it may not be desirable to let the data speak fully.
4.4. Uninformative Priors

Traditionally, most of the above descriptions of prior distributions were categorized as unin-
formative priors. However, uninformative priors do not truly exist (Irony and Singpurwalla
1997), and all priors are informative in some way. Traditionally, there have been many names
associated with uninformative priors, including diffuse, minimal, non-informative, objective,
reference, uniform, vague, and perhaps weakly informative.
4.5. Proper and Improper Priors

It is important for the prior distribution to be proper. A prior distribution, p(θ), is improper12
when
12
Improper priors were introduced in Jeffreys (1961).
Z
p(θ)dθ = ∞
As noted previously, an unbounded uniform prior distribution is an improper prior distribution

because p(θ) ∝ 1, for −∞ < θ < ∞. An improper prior distribution can cause an improper
posterior distribution. When the posterior distribution is improper, inferences are invalid, it
is non-integrable, and Bayes factors cannot be used (though there are exceptions).
To determine the propriety of a joint posterior distribution, the marginal likelihood must be
finite for all y. Again, the marginal likelihood is
Z
p(y) = p(y|Θ)p(Θ)dΘ
Although improper prior distributions can be used, it is good practice to avoid them.
5. Hierarchical Bayes
Prior distributions may be estimated within the model via hyperprior distributions, which are
usually vague and nearly flat. Parameters of hyperprior distributions are called hyperparam-
eters. Using hyperprior distributions to estimate prior distributions is known as hierarchical
Bayes. In theory, this process could continue further, using hyper-hyperprior distributions
to estimate the hyperprior distributions. Estimating priors through hyperpriors, and from
the data, is a method to elicit the optimal prior distributions. One of many natural uses for
hierarchical Bayes is multilevel modeling.
Recall that the unnormalized joint posterior distribution (equation 2) is proportional to the
likelihood times the prior distribution
p(Θ|y) ∝ p(y|Θ)p(Θ)
The simplest hierarchical Bayes model takes the form
p(Θ, Φ|y) ∝ p(y|Θ)p(Θ|Φ)p(Φ)
where Φ is a set of hyperprior distributions. By reading the equation from right to left, it
begins with hyperpriors Φ, which are used conditionally to estimate priors p(Θ|Φ), which
in turn is used, as per usual, to estimate the likelihood p(y|Θ), and finally the posterior is
p(Θ, Φ|y).
6. Conjugacy
When the posterior distribution p(Θ|y) is in the same family as the prior probability distri-
bution p(Θ), the prior and posterior are then called conjugate distributions, and the prior is
called a conjugate prior for the likelihood13 . For example, the Gaussian family is conjugate
to itself (or self-conjugate) with respect to a Gaussian likelihood function: if the likelihood
13
The conjugate prior approach was introduced in Raiffa and Schlaifer (1961).
Statisticat LLC 11
function is Gaussian, then choosing a Gaussian prior for the mean will ensure that the poste-
rior distribution is also Gaussian. All probability distributions in the exponential family have
conjugate priors. See Robert (2007) for a catalog.
Although the gamma distribution is the conjugate prior distribution for the precision of a
normal distribution (Spiegelhalter, Thomas, Best, and Lunn 2003),
τ ∼ G(0.001, 0.001),
better properties for scale parameters are yielded with the non-conjugate, proper, half-
Cauchy14 distribution, with a general recommendation of scale=25 for a weakly informative
scale parameter (Gelman 2006),
σ ∼ HC(25)
τ = σ −2
When the half-Cauchy is unavailable, a uniform distribution is often placed on σ in hierarchical

Bayes when the number of groups is, say, at least five,
σ ∼ U(0, 100)
τ = σ −2
Conjugacy is mathematically convenient in that the posterior distribution follows a known

parametric form (Gelman et al. 2004, p. 40). It is obviously easier to summarize a normal
distribution than a complex, multi-modal distribution with no known form. If information is
available that contradicts a conjugate parametric family, then it may be necessary to use a
more realistic, inconvenient, prior distribution.
The basic justification for the use of conjugate prior distributions is similar to that for using
standard models (such as the binomial and normal) for the likelihood: it is easy to understand
the results, which can often be put in analytic form, they are often a good approximation,
and they simplify computations. Also, they are useful as building blocks for more compli-
cated models, including many dimensions, where conjugacy is typically impossible. For these
reasons, conjugate models can be good starting points (Gelman et al. 2004, p. 41).
Nonconjugate prior distributions can make interpretations of posterior inferences less trans-
parent and computation more difficult, though this alternative does not pose any conceptual
problems. In practice, for complicated models, conjugate prior distributions may not even be
possible (Gelman et al. 2004, p. 41-42).
When conjugate distributions are used, a summary statistic for a posterior distribution of θ
may be represented as t(y) and said to be a sufficient statistic (Gelman et al. 2004, p. 42).
When nonconjugate distributions are used, a summary statistic for a posterior distribution is
usually not a sufficient statistic. A sufficient statistic is a statistic that has the property of
sufficiency with respect to a statistical model and the associated unknown parameter. The
quantity t(y) is said to be a sufficient statistic for θ, because the likelihood for θ depends
on the data y only through the value of t(y). Sufficient statistics are useful in algebraic
manipulations of likelihoods and posterior distributions.
14
The half-t distribution is another option.
7. Likelihood
In order to complete the definition of a Bayesian model, both the prior distributions and the
likelihood15 must be approximated or fully specified. The likelihood, likelihood function, or
p(y|Θ), contains the available information provided by the sample. The likelihood is
n
Y
p(y|Θ) = p(yi |Θ)
i=1
The data y affects the posterior distribution p(Θ|y) only through the likelihood function
p(y|Θ). In this way, Bayesian inference obeys the likelihood principle, which states that for
a given sample of data, any two probability models p(y|Θ) that have the same likelihood
function yield the same inference for Θ. For more information on the likelihood principle, see
section 7.2.
7.1. Terminology: From Inverse Probability to Bayesian Probability

A gambler’s dispute in 1654 led to the creation of a mathematical theory of probability by
two famous French mathematicians, Blaise Pascal and Pierre de Fermat. Reverend Thomas
Bayes (1701-1761) discovered Bayes’ theorem, published posthumously in 1763, in which he
was the first to use inverse probability (Bayes and Price 1763). ‘Inverse probability’ refers to
assigning a probability distribution to an unobserved variable, and is in essence, probability
in the opposite direction of the usual sense.
For example, the probability of obtaining heads on the next coin flip in a Bayesian context
would be the predicted probability, p(ynew |y, θ), but to estimate this predicted probability,
the probability distribution of θ must first be estimated, using coin toss data y to estimate
the parameter θ by the likelihood function p(y|θ), which contains the likelihood p(θ|y), where
θ is estimated from the data, y. Therefore, the data, y, is used to estimate the most probable
θ that would lead to a data-generating process for y.
Unaware of Bayes, Pierre-Simon Laplace (1749-1827) independently developed Bayes’ theorem
and first published his version in 1774, eleven years after Bayes, in one of Laplace’s first
major works (Laplace 1774, p. 366-367). In 1812, Laplace (1749-1827) introduced a host
of new ideas and mathematical techniques in his book, Theorie Analytique des Probabilites
(Laplace 1812). Before Laplace, probability theory was solely concerned with developing
a mathematical analysis of games of chance. Laplace applied probabilistic ideas to many
scientific and practical problems.
Then, in 1814, Laplace published his “Essai philosophique sur les probabilites”, which intro-
duced a mathematical system of inductive reasoning based on probability (Laplace 1814). In
it, the Bayesian interpretation of probability was developed independently by Laplace, much
15
Ronald A. Fisher, a prominent frequentist, introduced the term likelihood in 1921 (Fisher 1921), though
the concept of likelihood was used by Bayes and Laplace. Fisher’s introduction preceded a series of the most
influential papers in statistics (mostly in 1922 and 1925), in which Fisher introduced numerous terms that
are now common: consistency, efficiency, estimation, information, maximum likelihood estimate, optimality,
parameter, statistic, sufficiency, and variance. He was the first to use Greek letters for unknown parameters
and Latin letters for the estimates. Later contributions include F statistics, design of experiments, ANOVA,
and many more.
Statisticat LLC 13
more thoroughly than Bayes, so some “Bayesians” refer to Bayesian inference as Laplacian
inference.
The term “inverse probability” appears in an 1837 paper of Augustus De Morgan (De Morgan
1837), in reference to Laplace’s method of probability (Laplace 1774, 1812), though the term
“inverse probability” does not occur in these works. Bayes’ theorem has been referred to as
“the principle of inverse probability”.
Terminology has changed, so that today, Bayesian probability (rather than inverse probability)
refers to assigning a probability distribution to an unobservable variable. The “distribution”
of an unobserved variable given data is the likelihood function (which is not a distribution),
and the distribution of an unobserved variable, given both data and a prior distribution, is
the posterior distribution. The term “Bayesian”, which displaced “inverse probability”, was in
fact introduced by Ronald A. Fisher as a derogatory term.
In modern terms, given a probability distribution p(y|θ) for an observable quantity y con-
ditional on an unobserved variable θ, the “inverse probability” is the posterior distribution
p(θ|y), which depends both on the likelihood function (the inversion of the probability distri-
bution) and a prior distribution. The distribution p(y|θ) itself is called the direct probability.
However, p(y|θ) is also called the likelihood function, which can be confusing, seeming to pit
the definitions of probability and likelihood against each other. A quick introduction to the
likelihood principle follows, and finally all of the information on likelihood comes together in
the section entitled “Likelihood Function of a Parameterized Model”.
7.2. The Likelihood Principle

An informal summary of the likelihood principle may be that inferences from data to hypothe-
ses should depend on how likely the actual data are under competing hypotheses, not on how
likely imaginary data would have been under a single “null” hypothesis or any other properties
of merely possible data. Bayesian inferences depend only on the probabilities assigned due to
the observed data, not due to other data that might have been observed.
A more precise interpretation may be that inference procedures which make inferences about
simple hypotheses should not be justified by appealing to probabilities assigned to observations
that have not occurred. The usual interpretation is that any two probability models with the
same likelihood function yield the same inference for θ.
Some authors mistakenly claim that frequentist inference, such as the use of maximum like-
lihood estimation (MLE), obeys the likelihood, though it does not. Some authors claim that
the largest contention between Bayesians and frequentists regards prior probability distri-
butions. Other authors argue that, although the subject of priors gets more attention, the
true contention between frequentist and Bayesian inference is the likelihood principle, which
Bayesian inference obeys, and frequentist inference does not.
There have been many frequentist attacks on the likelihood principle, and have been shown to
be poor arguments. Some Bayesians have argued that Bayesian inference is incompatible with
the likelihood principle on the grounds that there is no such thing as an isolated likelihood
function (Bayarri and DeGroot 1987). They argue that in a Bayesian analysis there is no
principled distinction between the likelihood function and the prior probability function. The
objection is motivated, for Bayesians, by the fact that prior probabilities are needed in order
to apply what seems like the likelihood principle. Once it is admitted that there is a universal
necessity to use prior probabilities, there is no longer a need to separate the likelihood function
from the prior. Thus, the likelihood principle is accepted ‘conditional’ on the assumption that
a likelihood function has been specified, but it is denied that specifying a likelihood function
is necessary. Nonetheless, the likelihood principle is seen as a useful Bayesian weapon to
combat frequentism.
Following are some interesting qutoes from prominent statisticians:
“Using Bayes’ rule with a chosen probability model means that the data y affect
posterior inference ’only’ through the function p(y|θ), which, when regarded as a
function of θ, for fixed y, is called the ‘likelihood function’. In this way Bayesian
inference obeys what is sometimes called the ‘likelihood principle’, which states
that for a given sample of data, any two probability models p(y|θ) that have the
same likelihood function yield the same inference for θ” (Gelman et al. 2004, p. 9).
“The likelihood principle is reasonable, but only within the framework of the model
or family of models adopted for a particular analysis” (Gelman et al. 2004, p. 9).
Frequentist “procedures typically violate the likelihood principle, since long-run

behavior under hypothetical repetitions depends on the entire distribution p(y|θ),
y ∈ Y and not only on the likelihood” (Bernardo and Smith 2000, p. 454).
There is “a general fact about the mechanism of parametric Bayesian inference

which is trivially obvious; namely ‘for any specified p(θ), if the likelihood func-
tions p1 (y1 |θ), p2 (y2 |θ) are proportional as functions of θ, the resulting posterior
densities for θ are identical’. It turns out...that many non-Bayesian inference
procedures do not lead to identical inferences when applied to such proportional
likelihoods. The assertion that they ‘should’, the so-called ‘Likelihood Principle’,
is therefore a controversial issue among statisticians. In contrast, in the Bayesian
inference context...this is a straightforward consequence of Bayes’ theorem, rather
than an imposed ‘principle’ ” (Bernardo and Smith 2000, p. 249).
“Although the likelihood principle is implicit in Bayesian statistics, it was devel-

oped as a separate principle by Barnard (Barnard 1949), and became a focus of
interest when Birnbaum (1962) showed that it followed from the widely accepted
sufficiency and conditionality principles” (Bernardo and Smith 2000, p. 250).
“The likelihood principle, by itself, is not sufficient to build a method of inference

but should be regarded as a minimum requirement of any viable form of inference.
This is a controversial point of view for anyone familiar with modern economet-
rics literature. Much of this literature is devoted to methods that do not obey the
likelihood principle...” (Rossi, Allenby, and McCulloch 2005, p. 15).
“Adherence to the likelihood principle means that inferences are ‘conditional’ on

the observed data as the likelihood function is parameterized by the data. This
Statisticat LLC 15
is worth contrasting to any sampling-based approach to inference. In the sam-

pling literature, inference is conducted by examining the sampling distribution of
some estimator of θ, θ̂ = f (y). Some sort of sampling experiment results in a
distribution of y and therefore, the estimator is viewed as a random variable. The
sampling distribution of the estimator summarizes the properties of the estimator
‘prior’ to observing the data. As such, it is irrelevant to making inferences given
the data we actually observe. For any finite sample, this dinstinction is extremely
important. One must conclude that, given our goal for inference, sampling distri-
butions are simply not useful” (Rossi et al. 2005, p. 15).
7.3. Likelihood Function of a Parameterized Model

In non-technical parlance, “likelihood” is usually a synonym for “probability”, but in statistical
usage there is a clear distinction: whereas “probability” allows us to predict unknown outcomes
based on known parameters, “likelihood” allows us to estimate unknown parameters based on
known outcomes.
In a sense, likelihood can be thought a reversed version of conditional probability. Reasoning
forward from a given parameter θ, the conditional probability of y is the density p(y|θ). With
θ as a parameter, here are relationships in expressions of the likelihood function
L (θ|y) = p(y|θ) = f (y|θ)
where y is the observed outcome of an experiment, and the likelihood (L ) of θ given y is

equal to the density p(y|θ) or function f (y|θ). When viewed as a function of y with θ fixed, it
is not a likelihood function L (θ|y), but merely a probability density function p(y|θ). When
viewed as a function of θ with y fixed, it is a likelihood function and may be denoted as
L (θ|y), p(y|θ), or f (y|θ)16 .
For example, in a Bayesian linear regression with an intercept and two independent variables,
the model may be specified as
yi ∼ N (µi , σ 2 )
µi = β1 + β2 Xi,1 + β3 Xi,2
The dependent variable y, indexed by i = 1, ..., n, is stochastic, and normally-distributed

according to the expectation vector µ, and variance σ 2 . Expectation vector µ is an additive,
linear function of a vector of regression parameters, β, and the design matrix X.
Since y is normally-distributed, the probability density function (PDF) of a normal distribu-
tion will be used, and is usually denoted as
1 1
f (y) = √ exp[(− σ 2 )(yi − µi )2 ]; y ∈ (−∞, ∞)
2πσ 2
By considering a conditional distribution, the record-level likelihood in Bayesian notation is
16
Note that L (θ|y) is not the same as the probability that those parameters are the right ones, given the
observed sample.
1 1
p(yi |Θ) = √ exp[(− σ 2 )(yi − µi )2 ]; y ∈ (−∞, ∞)
2πσ 2
In both theory and practice, and in both frequentist and Bayesian inference, the log-likelihood
is used instead of the likelihood, on both the record- and model-level. The model-level product
of record-level likelihoods can exceed the range of a number that can be stored by a computer,
which is usually affected by sample size. By estimating a record-level log-likelihood, rather
than likelihood, the model-level log-likelihood is the sum of the record-level log-likelihoods,
rather than a product of the record-level likelihoods.
n
X
log[p(y|θ)] = log[p(yi |θ)]
i=1
rather than
n
Y
p(y|θ) = p(yi |θ)
i=1
As a function of θ, the unnormalized joint posterior distribution is the product of the likeli-
hood function and the prior distributions. To continue with the example of Bayesian linear
regression, here is the unnormalized joint posterior distribution
p(β, σ 2 |y) = p(y|β, σ 2 )p(β1 )p(β2 )p(β3 )p(σ 2 )
More usually, the logarithm of the unnormalized joint posterior distribution is used, which is
the sum of the log-likelihood and prior distributions. Here is the logarithm of the unnormalized
joint posterior distribution for this example
log[p(β, σ 2 |y)] = log[p(y|β, σ 2 )] + log[p(β1 )] + log[p(β2 )] + log[p(β3 )] + log[p(σ 2 )]
The logarithm of the unnormalized joint posterior distribution is maximized with numerical
approximation.
8. Numerical Approximation
The technical problem of evaluating quantities required for Bayesian inference typically re-
duces to the calculation of a ratio of two integrals (Bernardo and Smith 2000, p. 339). In all
cases, the technical key to the implementation of the formal solution given by Bayes’ theorem
is the ability to perform a number of integrations (Bernardo and Smith 2000, p. 340). Except
in certain rather stylized problems, the required integrations will not be feasible analytically
and, thus, efficient approximation strategies are required.
There are too many different types of numerical approximation algorithms in Bayesian infer-
ence to cover in any detail in this article. An incomplete list of broad categories of Bayesian
numerical approximation may include Approximate Bayesian Computation (ABC), Impor-
tance Sampling, Iterative Quadrature, Laplace Approximation, Markov chain Monte Carlo
Statisticat LLC 17
(MCMC), and Variational Bayes (VB). For more information on algorithms in LaplacesDe-
mon, see the accompanying vignette entitled “LaplacesDemon Tutorial”.
Approximate Bayesian Computation (ABC), also called likelihood-free estimation, is a family
of numerical approximation techniques in Bayesian inference. ABC is especially useful when
evaluation of the likelihood, p(y|Θ) is computationally prohibitive, or when suitable likeli-
hoods are unavailable. As such, ABC algorithms estimate likelihood-free approximations.
ABC is usually faster than a similar likelihood-based numerical approximation technique,
because the likelihood is not evaluated directly, but replaced with an approximation that is
usually easier to calculate. The approximation of a likelihood is usually estimated with a
measure of distance between the observed sample, y, and its replicate given the model, yrep ,
or with summary statistics of the observed and replicated samples.
Importance Sampling is a method of estimating a distribution with samples from a different
distribution, called the importance distribution. Importance weights are assigned to each
sample. The main difficulty with importance sampling is in the selection of the importance
distribution. Importance sampling is the basis of a wide variety of algorithms, some of which
involve the combination of importance sampling and Markov chain Monte Carlo. There are
also many variations of importance sampling, including adaptive importance sampling, and
parametric and nonparametric self-normalized importance sampling. Population Monte Carlo
(PMC) is based on adaptive importance sampling.
Iterative quadrature is a traditional approach to evaluating integrals. Multidimensional
quadrature, often called cubature, performs well, but is limited usually to ten or fewer param-
eters. Componentwise quadrature may be applied to any model regardless of dimension, but
estimates only variance, rather than covariance. Bayesian quadrature typically uses adaptive
Gauss-Hermite quadrature, which assumes the marginal posterior distributions are normally-
distrubted. Under this assumption, the conditional mean and conditional variance of each
distribution is adapted each iteration according to the evaluation of samples determined by
quadrature rules.
Laplace Approximation dates back to Laplace (1774, 1814), and is used to approximate the
posterior moments of integrals. Specifically, the posterior mode is estimated for each param-
eter, assumed to be unimodal and Gaussian. As a Gaussian distribution, the posterior mean
is the same as the posterior mode, and the variance is estimated. Laplace Approximation is
a family of deterministic algorithms that usually converge faster than MCMC, and just a lit-
tle slower than Maximum Likelihood Estimation (MLE) (Azevedo-Filho and Shachter 1994).
Laplace Approximation shares many limitations of MLE, including asymptotic estimation
with respect to sample size.
MCMC algorithms originated in statistical physics and are now used in Bayesian inference to
sample from probability distributions by constructing Markov chains. In Bayesian inference,
the target distribution of each Markov chain is usually a marginal posterior distribution, such
as each parameter θ. Each Markov chain begins with an initial value and the algorithm iter-
ates, attempting to maximize the logarithm of the unnormalized joint posterior distribution
and eventually arriving at each target distribution. Each iteration is considered a state. A
Markov chain is a random process with a finite state-space and the Markov property, meaning
that the next state depends only on the current state, not on the past. The quality of the
marginal samples usually improves with the number of iterations.
A Monte Carlo method is an algorithm that relies on repeated pseudo-random sampling for
computation, and is therefore stochastic (as opposed to deterministic). Monte Carlo methods
are often used for simulation. The union of Markov chains and Monte Carlo methods is
called MCMC. The revival of Bayesian inference since the 1980s is due to MCMC algorithms
and increased computing power. The most prevalent MCMC algorithms may be the simplest:
random-walk Metropolis and Gibbs sampling. There are a large number of MCMC algorithms,
and further details on MCMC are best explored outside of this article.
VB is a family of algorithms within variational inference. VB are deterministic optimization
algorithms that approximate the posterior with a distribution. Each marginal posterior dis-
tribution is estimated with an approximating distribution. VB usually converges faster than
MCMC. VB shares many limitations of MLE, including asymptotic estimation with respect
to sample size.
9. Prediction
The “posterior predictive distribution” is either the replication of y given the model (usually
represented as yrep ), or the prediction of a new and unobserved y (usually represented as ynew
or y′ ), given the model. This is the likelihood of the replicated or predicted data, averaged
over the posterior distribution p(Θ|y)
Z
rep
p(y |y) = p(yrep |Θ)p(Θ|y)dΘ
or
Z
p(ynew |y) = p(ynew |Θ)p(Θ|y)dΘ
If y has missing values, then the missing ys can be estimated with the posterior predictive
distribution17 as ynew from within the model. For the linear regression example, the integral
for prediction is
Z
p(ynew |y) = p(ynew |β, σ 2 )p(β, σ 2 |y)dβdσ 2
The posterior predictive distribution is easy to estimate
ynew ∼ N (µ, σ 2 )
where µ = Xβ, and µ is the conditional mean, while σ 2 is the residual variance.
10. Bayes Factors

Introduced by Harold Jeffreys, a ‘Bayes factor’ is a Bayesian alternative to frequentist hy-
pothesis testing that is most often used for the comparison of multiple models by hypothesis
testing, usually to determine which model better fits the data (Jeffreys 1961). Bayes factors
are notoriously difficult to compute, and the Bayes factor is only defined when the marginal
17
The predictive distribution was introduced by Jeffreys (1961).
Statisticat LLC 19
density of y under each model is proper. However, Bayes factors are easy to approximate
with the Laplace-Metropolis Estimator (Kass and Raftery 1995; Lewis and Raftery 1997)18 .
Hypothesis testing with Bayes factors is more robust than frequentist hypothesis testing, since
the Bayesian form avoids model selection bias, evaluates evidence in favor the null hypothesis,
includes model uncertainty, and allows non-nested models to be compared (though of course
the model must have the same dependent variable). Also, frequentist significance tests become
biased in favor of rejecting the null hypothesis with sufficiently large sample size.
The Bayes factor for comparing two models may be approximated as the ratio of the marginal
likelihood of the data in model 1 and model 2. Formally, the Bayes factor in this case is
R
p(y|M1 ) p(y|Θ1 , M1 )p(Θ1 |M1 )dΘ1
B= =R
p(y|M2 ) p(y|Θ2 , M2 )p(Θ2 |M2 )dΘ2
where p(y|M1 ) is the marginal likelihood of the data in model 1.

The Bayes factor, B, is the posterior odds in favor of the hypothesis divided by the prior odds
in favor of the hypothesis, where the hypothesis is usually M1 > M2 . Put another way,
(Posterior model odds) = (Bayes factor) x (prior model odds)
For example, when B = 2, the data favor M1 over M2 with 2:1 odds.
In a non-hierarchical model, the marginal likelihood may easily be approximated with the
Laplace-Metropolis Estimator for model m as
p(y|m) = (2π)dm /2 |Σm |1/2 p(y|Θm , m)p(Θm |m)
where d is the number of parameters and Σ is the inverse of the negative of the Hessian matrix
of second derivatives. Lewis and Raftery (1997) introduce the Laplace-Metropolis method of
approximating the marginal likelihood in MCMC, though it naturally works with Laplace
Approximation as well. For a hierarchical model that involves both fixed and random effects,
the Compound Laplace-Metropolis Estimator must be used.
Gelman finds Bayes factors generally to be irrelevant, because they compute the relative
probabilities of the models conditional on one of them being true. Gelman prefers approaches
that measure the distance of the data to each of the approximate models (Gelman et al.
2004, p. 180). However, Kass and Raftery (1995) explain that “the logarithm of the marginal
probability of the data may also be viewed as a predictive score. This is of interest, because
it leads to an interpretation of the Bayes factor that does not depend on viewing one of the
models as ‘true”’.
Three of many possible alternatives are to use
1. pseudo Bayes factors (PsBF) based on a ratio of pseudo marginal likelihoods (PsMLs)
2. Deviance Information Criterion (DIC)

18
A Bayes factor may be estimated with the BayesFactor function in LaplacesDemon to compare multiple
models that were fit with the LaplaceApproximation or LaplacesDemon functions. See the BayesFactor
function for the interpretation of a Bayes factor regarding strength of evidence.
3. Widely Applicable Information Criterion (WAIC)
DIC is the most popular method of assessing model fit and comparing models, though Bayes
factors are better, when appropriate, because they take more into account. WAIC is a newer
criterion.
11. Model Fit

In Bayesian inference, the most common method of assessing the goodness of fit of an es-
timated statistical model is a generalization of the frequentist Akaike Information Criterion
(AIC). The Bayesian method, like AIC, is not a test of the model in the sense of hypothesis
testing, though Bayesian inference has Bayes factors for such purposes. Instead, like AIC,
Bayesian inference provides a model fit statistic that is to be used as a tool to refine the
current model or select the better-fitting model of different methodologies.
To begin with, model fit can be summarized with deviance, which is defined as -2 times the
log-likelihood (Gelman et al. 2004, p. 180), such as
D(y, Θ) = −2 log[p(y|Θ)]
Just as with the likelihood, p(y|Θ), or log-likelihood, the deviance exists at both the record-
and model-level. Due to the development of BUGS software (Gilks, Thomas, and Spiegelhalter
1994), deviance is defined differently in Bayesian inference than frequentist inference. In
frequentist inference, deviance is -2 times the log-likelihood ratio of a reduced model compared
to a full model, whereas in Bayesian inference, deviance is simply -2 times the log-likelihood.
In Bayesian inference, the lowest expected deviance has the highest posterior probability
(Gelman et al. 2004, p. 181).
It is possible to have a negative deviance. Deviance is derived from the likelihood, which is
derived from probability density functions (PDF). Evaluated at a certain point in parameter
space, a PDF can have a density larger than 1 due to a small standard deviation or lack of
variation. Likelihoods greater than 1 lead to negative deviance, and are appropriate.
On its own, the deviance is an insufficient model fit statistic, because it does not take model
complexity into account. The effect of model fitting, pD, is used as the ‘effective number of
parameters’ of a Bayesian model. The sum of the differences between the posterior mean of
the model-level deviance and the deviance at each draw i of θi is the pD.
A related way to measure model complexity is as half the posterior variance of the model-level
deviance, known as pV (Gelman et al. 2004, p. 182)
pV = var(D)/2
The effect of model fitting, pD or pV, can be thought of as the number of ‘unconstrained’
parameters in the model, where a parameter counts as: 1 if it is estimated with no constraints
or prior information; 0 if it is fully constrained or if all the information about the parameter
comes from the prior distribution; or an intermediate value if both the data and the prior are
informative (Gelman et al. 2004, p. 182). Therefore, by including prior information, Bayesian
inference is more efficient in terms of the effective number of parameters than frequentist
Statisticat LLC 21
inference. Hierarchical, mixed effects, or multilevel models are even more efficient regarding
the effective number of parameters.
Model complexity, pD or pV, should be positive. Although pV must be positive since it is
related to variance, it is possible for pD to be negative, which indicates one or more problems:
log-likelihood is non-concave, a conflict between the prior and the data, or that the posterior
mean is a poor estimator (such as with a bimodal posterior).
The sum of both the mean model-level deviance and the model complexity (pD or pV) is
the Deviance Information Criterion (DIC), a model fit statistic that is also an estimate of
the expected loss, with deviance as a loss function (Spiegelhalter, Best, and Carlin 1998;
Spiegelhalter, Best, Carlin, and van der Linde 2002). DIC is
DIC = D̄ + pV
DIC may be compared across different models and even different methods, as long as the
dependent variable does not change between models, making DIC the most flexible model fit
statistic. DIC is a hierarchical modeling generalization of the Akaike Information Criterion
(AIC) and Bayesian Information Criterion (BIC). Like AIC and BIC, it is an asymptotic
approximation as the sample size becomes large. DIC is valid only when the joint posterior
distribution is approximately multivariate normal. Models should be preferred with smaller
DIC. Since DIC increases with model complexity (pD or pV), simpler models are preferred.
It is difficult to say what would constitute an important difference in DIC. Very roughly,
differences of more than 10 might rule out the model with the higher DIC, differences between
5 and 10 are substantial, but if the difference in DIC is, say, less than 5, and the models make
very different inferences, then it could be misleading just to report the model with the lowest
DIC.
The Widely Applicable Information Criterion (WAIC) is an information criterion that is
more fully Bayesian than DIC. WAIC is more difficult to calculate because the record-level
log-likelihood is required over numerous samples. However, when available, the result more
closely resembles leave-one-out cross-validation (LOO-CV).
The Bayesian Predictive Information Criterion (BPIC) was introduced as a criterion of model
fit when the goal is to pick a model with the best out-of-sample predictive power (Ando 2007).
BPIC is a variation of DIC where the effective number of parameters is 2pD (or 2pV). BPIC
may be compared between ynew and yholdout , and has many other extensions, such as with
Bayesian Model Averaging (BMA).
12. Posterior Predictive Checks

Comparing the predictive distribution yrep to the observed data y is generally termed a
“posterior predictive check”. This type of check includes the uncertainty associated with the
estimated parameters of the model, unlike frequentist statistics.
Posterior predictive checks (via the predictive distribution) involve a double-use of the data,
which violates the likelihood principle. However, arguments have been made in favor of
posterior predictive checks, provided that usage is limited to measures of discrepancy to
study model adequacy, not for model comparison and inference (Meng 1994).
Gelman recommends at the most basic level to compare yrep to y, looking for any systematic
differences, which could indicate potential failings of the model (Gelman et al. 2004, p. 159).
It is often first recommended to compare graphical plots, such as the distribution of y and
yrep . There are many posterior predictive checks that are not included in this article, but an
introduction to a selection of them appears below.
12.1. Bayesian p-values

A Bayesian form of p-value may be estimated with a variety of test statistics (Gelman, Meng,
and Stern 1996). Usually the minimum or maximum observed y is compared to the minimum
or maximum yrep . A Bayesian p-value is one of several ways to report discrepancies between
y and yrep .
Frequentist p-values have many problems, but here it will only be noted that the frequen-
tist p-value estimates p(data|hypothesis), while in this case the Bayesian form estimates
p(hypothesis|data). The frequentist estimates the wrong probability, because the frequen-
tist is forced to consider the parameters to be fixed and the data random, projecting long-run
frequencies of what should happen with future, repeated sampling of similar data, given a
fixed parameter, or in this case hypothesis. Even the term hypothesis testing suggests you
want to test the hypothesis given the data, not the data given the hypothesis19 .
12.2. Chi-Square
Gelman et al. (2004, p. 175) suggest an omnibus test such as the following χ2
PT
yrep
(y − t=1T i,t )2
χ2i = i ,
var(yrep
i,1:T )
over records i = 1, . . . , N and posterior samples t = 1, . . . , T . The sum of χ2i over records
i = 1, . . . , N is an overall goodness of fit measure on the data set. Larger values of χ2i indicate
a worse fit for each record.
An alternative χ2 test is
p(χ2rep 2obs
i,1:T > χi,1:T )
where a worse fit is indicated as p approaches zero or one, and it is common to consider
records with a poor fit to be outside the 95% probability interval. To continue
[yi − E(yi )]2

χ2obs
i,1:T =
E(yi )
and
[yrep rep 2
i,1:T − E(yi )]
χ2rep
i,1:T =
E(yrep
i )
19
Numerous problems with frequentist p-values, confidence intervals, point estimates, and hypothesis testing
are worth exploring, but not detailed in this article.
Statisticat LLC 23
Newer forms of χ2 tests have been proposed in the literature, and are best explored elsewhere.
12.3. Conditional Predictive Ordinate

Although the full predictive distribution p(yrep |y) is useful for prediction, its use for model-
checking is questionable because of the double-use of the data, and causes predictive perfor-
mance to be overestimated. The leave-one-out cross-validation predictive density has been
proposed (Geisser and Eddy 1979). This is also known as the Conditional Predictive Ordinate
or CPO (Gelfand 1996).
The CPO is
Z
p(yi |y[i] ) = p(yi |Θ)p(Θ|y[i] )dΘ
where yi is each instance of an observed y, and y[i] is y without the current observation i.
The CPO is easy to calculate with MCMC or PMC numerical approximation. By considering
the inverse likelihood across T iterations, the CPO for each individual i is
1
CPOi = T
X
T −1 p(yi |Θt )−1
t=1
The CPO is a handy posterior predictive check because it may be used to identify outliers, in-
fluential observations, and for hypothesis testing across different non-nested models. However,
it may be difficult to calculate with latent mixtures.
The CPO expresses the posterior probability of observing the value (or set of values) of yi
when the model is fitted to all data except yi , with a larger value implying a better fit of the
model to yi , and very low CPO values suggest that yi is an outlier and an influential obser-
vation. A Monte Carlo estimate of the CPO is obtained without actually omitting yi from
the estimation, and is provided by the harmonic mean of the likelihood for yi . Specifically,
the CP Oi is the inverse of the posterior mean of the inverse likelihood of yi .
The CPO is connected with the frequentist studentized residual test for outlier detection.
Data with large studentized residuals have small CPOs and will be detected as outliers.
An advantage of the CPO is that observations with high leverage will have small CPOs,
independently of whether or not they are outliers. The Bayesian CPO is able to detect both
outliers and influential points, whereas the frequentist studentized residual is unable to detect
high-leverage outliers.
Inverse-CPOs (ICPOs) larger than 40 can be considered as possible outliers, and higher than
70 as extreme values (Ntzoufras 2009, p. 376). Congdon recommends scaling CPOs by divid-
ing each by its individual maximum (after the posterior mean) and considering observations
with scaled CPOs under 0.01 to be outliers (Congdon 2005). The range in scaled CPOs is
useful as an indicator of a good-fitting model.
The sum of the logged CPOs can be an estimator for the logarithm of the marginal likeli-
hood20 , sometimes called the log pseudo marginal likelihood (LPsML). A ratio of PsMLs is
20
Exercise extreme caution when approximating the marginal likelihood from CPOs, or use a method with
better repute, such as the Laplace-Metropolis Estimator or importance sampling.
a surrogate for the Bayes factor, sometimes known as the pseudo Bayes factor (PsBF). In
this way, non-nested models may be compared with a hypothesis test to determine the better
model, if one exists, based on leave-one-out cross-validation.
12.4. Predictive Concordance

Gelfand (1996) suggests that any yi that is in either 2.5% tail area of yrep
i should be considered
an outlier. For each i, I am calling this the predictive quantile (PQ), which is calculated as
PQi = p(yrep
i > yi )
and is somewhat similar to the Bayesian p-value. The percentage of yi s that are not outliers
is called the ‘Predictive Concordance’. Gelfand (1996) suggests the goal is to attempt to
achieve 95% predictive concordance. In the case of, say 80% predictive concordance, the
discrepancy between the model and data is undesirable because the model does not fit the
data well and many outliers have resulted. On the other hand, if the predictive concordance
is too high, say 100%, then overfitting may have occurred, and it may be worth considering
a more parsimonious model. Kernel density plots of each yrep i distribution are useful in this
case with the actual yi included as a vertical bar to show its position.
12.5. L-criterion
Laud and Ibrahim (1995) introduced the L-criterion as one of three posterior predictive checks
for model, variable, and transformation selection. The L-criterion is a posterior predictive
check that is widely applicable and easy to apply. It is a sum of two components: one involves
the predictive variance and the other includes the accuracy of the means of the predictive
distribution. The L-criterion measures model performance with a combination of how close
its predictions are to the observed data and variability of the predictions. Better models have
smaller values of L. L is measured in the same units as the response variable, and measures
how close the data vector y is to the predictive distribution. In addition to the value of L,
there is a value for SL , which is the calibration number of L, and is useful in determining how
much of a decrease is necessary between models to be noteworthy. The L-criterion is
s
N PT rep
t=1 yi,t 2
var(yrep
X
L= i,1:T ) + (yi − ) ,
T
i=1
over t = 1, . . . , T posterior samples. The calibration number, SL , is the standard deviation of

L over records i = 1, . . . , N .
Gelfand and Ghosh (1998) introduced Posterior Predictive Loss (PPL). This posterior pre-
dictive check for model comparison may be viewed as an extension to the L-criterion in which
a weight k is applied to the accuracy (fit) component.
13. Advantages Of Bayesian Inference Over Frequentist Inference

Following is a short list of advantages of Bayesian inference over frequentist inference.
Statisticat LLC 25
• Bayesian inference allows informative priors so that prior knowledge or results of a

previous model can be used to inform the current model.
• Bayesian inference can avoid problems with model identification by manipulating prior
distributions (usually in complex models). Frequentist inference with any numerical
approximation algorithm does not have prior distributions, and can become stuck in
regions of flat density, causing problems with model identification.
• Bayesian inference considers the data to be fixed (which it is), and parameters to be
random because they are unknowns. Frequentist inference considers the unknown pa-
rameters to be fixed, and the data to be random, estimating not based on the data
at hand, but the data at hand plus hypothetical repeated sampling in the future with
similar data. “The Bayesian approach delivers the answer to the right question in the
sense that Bayesian inference provides answers conditional on the observed data and
not based on the distribution of estimators or test statistics over imaginary samples not
observed” (Rossi et al. 2005, p. 4).
• Bayesian inference estimates a full probability model. Frequentist inference does not.
There is no frequentist probability distribution associated with parameters or hypothe-
ses.
• Bayesian inference estimates p(hypothesis|data). In contrast, frequentist inference esti-

mates p(data|hypothesis). Even the term ’hypothesis testing’ suggests it should be the
hypothesis that is tested, given the data, not the other way around.
• Bayesian inference has an axiomatic foundation (Cox 1946) that is uncontested by fre-
quentists. Therefore, Bayesian inference is coherent to a frequentist, but frequentist
inference is incoherent to a Bayesian.
• Bayesian inference has a decision theoretic foundation (Bernardo and Smith 2000;
Robert 2007). The purpose of most of statistical inference is to facilitate decision-
making (Robert 2007, p. 51). The optimal decision is the Bayesian decision.
• Bayesian inference includes uncertainty in the probability model, yielding more real-
istic predictions. Frequentist inference does not include uncertainty of the parameter
estimates, yielding less realistic predictions.
• Bayesian inference is consistent with much of philosophy of science regarding epistemol-

ogy, where knowledge cannot be built entirely through experimentation, but requires
prior knowledge (Robert 2007, p. 510). Elsewhere, it has been suggested that the best
choice for philosophy of science is through Bayesian inference.
• Bayesian inference may use DIC to compare models with different methods includ-
ing hierarchical models, where frequentist model fit statistics cannot compare different
methods or hierarchical models.
• Bayesian inference obeys the likelihood principle. Frequentist inference, including Max-
imum Likelihood Estimation (MLE) and the General Method of Moments (GMM) or
Generalized Estimating Equations (GEE), violates the likelihood principle. “The likeli-
hood principle, by itself, is not sufficient to build a method of inference but should be
regarded as a minimum requirement of any viable form of inference. This is a contro-

versial point of view for anyone familiar with modern econometrics literature. Much of
this literature is devoted to methods that do not obey the likelihood principle...” (Rossi
et al. 2005, p. 15).
• Bayesian inference safeguards against overfitting by integrating over model parameters.
While Bayesian inference is not immune to overfitting, overfitting is largely a frequentist
problem.
• Bayesian inference uses observed data only. Frequentist inference uses both observed
data and future data that is unobserved and hypothetical.
• Bayesian inference uses prior distributions, so more information is used and 95% prob-
ability intervals of posterior distributions should be narrower than 95% confidence in-
tervals of frequentist point-estimates.
• Bayesian inference uses probability intervals (quantile-based, highest posterior density,
or preferably lowest posterior loss) to state the probability that θ is between two points.
Frequentist inference uses confidence intervals, which must be interpreted with proba-
bility of zero or one that θ is in the region, and the frequentist never knows whether
it is or is not, but can only say that if 100 repeated samples were drawn in the future,
that it would be in the region for 95 samples.
• Bayesian inference via MCMC or PMC algorithms allows more complicated models that
frequentists are unable to estimate.
• Bayesian inference via MCMC has a theoretic guarantee that the MCMC algorithm will
converge if run long enough. Frequentist inference with Maximum Likelihood Estima-
tion (MLE) has no guarantee of convergence.
• Bayesian inference via MCMC or PMC is unbiased with respect to sample size and
can accommodate any sample size no matter how small. Frequentist inference becomes
more biased as sample size decreases from infinity, and is often wildly biased with small
samples, so minimum sample size is an issue. Conversely, frequentist inference with
large sample sizes biases p-values to indicate that insignificant effects are significant.
• Bayesian inference via MCMC or PMC uses exact estimation with respect to sample
size. Frequentist inference uses approximate estimation that relies on asymptotic theory.
• Bayesian inference with correlated predictors sometimes allows the hyperparameters to
be distributed multivariate-normal, therefore including such correlation into the MCMC
or PMC algorithm to improve estimation. Frequentist inference does not use prior dis-
tributions, so confidence intervals are wider and less certain with correlated predictors.
• Bayesian inference with proper priors is immune to singularities and near-singularities
with matrix inversions, unlike frequentist inference.
14. Advantages Of Frequentist Inference Over Bayesian Inference

Following is a short list of advantages of frequentist inference over Bayesian inference.
Statisticat LLC 27
• Frequentist models are perceived to handle large data sets, while Bayesian models via
MCMC have traditionally been restricted to small sample sizes, and Laplace Approxi-
mation is similar to the frequentist method in that it is known to be able to handle large
data sets. This reputation is no longer true for MCMC. Samplers in LaplacesDemon do
not usually loop through records and can handle large data sets. But most importantly,
algorithms now exist (and are available here) that enable fast Bayesian inference with
big data.
• Frequentist models are usually much easier to prepare because many things do not need
to be specified, such as prior distributions, initial values for numerical approximation,
and usually the likelihood function. Most frequentist methods have been standardized
to “procedures” where less knowledge and programming are required, and in many cases
the user can just click on a few things and not really know what they are doing. Garbage
in, garbage out.
• Frequentist models optimized with MLE have much shorter run-times than Bayesian
models via MCMC or PMC. This is not a difference between frequentist and Bayesian
methods, but is due to optimization vs. sampling algorithms. MCMC has a longer run-
time, whether it is Bayesian or frequentist. Laplace Approximation uses optimization
algorithms, and yields run-times similar to frequentist MLE. If frequentist MLE and
Bayesian Laplace Approximation seem to have very different run-times, it is probably
due to differences between a method-specific algorithm vs. a general-purpose algorithm.
As they say, it pays to go Bayes.
References
Ando T (2007). “Bayesian Predictive Information Criterion for the Evaluation of Hierarchical
Bayesian and Empirical Bayes Models.” Biometrika, 94(2), 443–458.
Anscombe F, Aumann R (1963). “A Definition of Subjective Probability.” The Annals of
Mathematical Statistics, 34(1), 199–205.
Azevedo-Filho A, Shachter R (1994). “Laplace’s Method Approximations for Probabilistic
Inference in Belief Networks with Continuous Variables.” In R Mantaras, D Poole (eds.),
Uncertainty in Artificial Intelligence, pp. 28–36. Morgan Kauffman, San Francisco, CA.
Barnard G (1949). “Statistical Inference.” Journal of the Royal Statistical Society, B 11,
115–149.
Bayarri M, DeGroot M (1987). “Bayesian Analysis of Selection Models.” The Statistician,
36, 136–146.
Bayes T, Price R (1763). “An Essay Towards Solving a Problem in the Doctrine of Chances.
By the late Rev. Mr. Bayes, communicated by Mr. Price, in a letter to John Canton, MA.
and F.R.S.” Philosophical Transactions of the Royal Society of London, 53, 370–418.
Berger J (2006). “The Case for Objective Bayesian Analysis.” Bayesian Analysis, 1(3), 385–
402.
Berger J, Bernardo J, Dongchu S (2009). “The Formal Definition of Reference Priors.” Annals
of Statistics, 37(2), 905–938.
Bernardo J (1979). “Reference Posterior Distributions for Bayesian Inference (with discus-
sion).” Journal of the Royal Statistical Society, B 41, 113–147.
Bernardo J (2005). “Reference Analysis.” In D Dey, C Rao (eds.), Handbook of Statistics 25,
pp. 17–90. Elsevier, Amsterdam.
Bernardo J (2008). “Comment on Article by Gelman.” Bayesian Analysis, 3(3), 451–454.
Bernardo J, Smith A (2000). Bayesian Theory. John Wiley & Sons, West Sussex, England.
Congdon P (2005). Bayesian Models for Categorical Data. John Wiley & Sons, West Sussex,
England.
Cox R (1946). “Probability, Frequency and Reasonable Expectation.” American Journal of

Physics, 14(1), 1–13.
De Finetti B (1931). “Probabilismo.” Erkenntnis, 31, 169–223. English translation as “Prob-

abilism: A Critical Essay on the Theory of Probability and on the Value of Science”.
De Finetti B (1937). “La Prevision: ses lois logigues, ses sources subjectives.” Annales de
l’Institut Henri Poincare. English translation in H.E. Kyburg and H.E. Smokler (eds),
(1964), “Foresight: Its Logical Laws, Its Subjective Sources”, Studies in Subjective Proba-
bility, New York: Wiley.
De Morgan A (1837). “Review of Laplace’s Theorie Analytique des Probabilites.” Dublin

Review, 2,3, 338–354,237–354.
Fisher R (1921). “On the ‘Probable Error’ of a Coefficient of Correlation Deduced From a
Small Sample.” Metron, 1(4), 3–32.
Geisser S, Eddy W (1979). “A Predictive Approach to Model Selection.” Journal of the

American Statistical Association, 74, 153–160.
Gelfand A (1996). “Model Determination Using Sampling Based Methods.” In W Gilks,

S Richardson, D Spiegelhalter (eds.), Markov Chain Monte Carlo in Practice, pp. 145–161.
Chapman & Hall, Boca Raton, FL.
Gelfand A, Ghosh S (1998). “Model Choice: A Minimum Posterior Predictive Loss Approach.”
Biometrika, 85, 1–11.
Gelman A (2006). “Prior Distributions for Variance Parameters in Hierarchical Models.”

Bayesian Analysis, 1(3), 515–533.
Gelman A (2008). “Scaling Regression Inputs by Dividing by Two Standard Deviations.”

Statistics in Medicine, 27, 2865–2873.
Gelman A, Carlin J, Stern H, Rubin D (2004). Bayesian Data Analysis. 2nd edition. Chapman
& Hall, Boca Raton, FL.
Statisticat LLC 29
Gelman A, Meng X, Stern H (1996). “Posterior Predictive Assessment of Model Fitness via
Realized Discrepancies.” Statistica Sinica, 6, 773–807.
Gilks W, Thomas A, Spiegelhalter D (1994). “A Language and Program for Complex Bayesian
Modelling.” The Statistician, 43(1), 169–177.
Goldstein M (2006). “Subjective Bayesian Analysis: Principles and Practice.” Bayesian

Analysis, 1(3), 403–420.
Ibrahim J, Chen M (2000). “Power Prior Distributions for Regression Models.” Statistical
Science, 15, 46–60.
Irony T, Singpurwalla N (1997). “Noninformative Priors Do Not Exist: a Discussion with

Jose M. Bernardo.” Journal of Statistical Inference and Planning, 65, 159–189.
Jaynes E (1968). “Prior Probabilities.” IEEE Transactions on Systems Science and Cyber-
netics, 4(3), 227–241.
Jeffreys H (1961). Theory of Probability. Third edition. Oxford University Press, Oxford,
England.
Kass R, Raftery A (1995). “Bayes Factors.” Journal of the American Statistical Association,
90(430), 773–795.
Lambert P, Sutton A, Burton P, Abrams K, Jones D (2005). “How Vague is Vague? A

Simulation Study of the Impact of the Use of Vague Prior Distributions in MCMC using
WinBUGS.” Statistics in Medicine, 24, 2401–2428.
Laplace P (1774). “Memoire sur la Probabilite des Causes par les Evenements.” l’Academie
Royale des Sciences, 6, 621–656. English translation by S.M. Stigler in 1986 as “Memoir
on the Probability of the Causes of Events” in Statistical Science, 1(3), 359–378.
Laplace P (1812). Theorie Analytique des Probabilites. Courcier, Paris. Reprinted as “Oeuvres
Completes de Laplace”, 7, 1878–1912. Paris: Gauthier-Villars.
Laplace P (1814). “Essai Philosophique sur les Probabilites.” English translation in Truscott,
F.W. and Emory, F.L. (2007) from (1902) as “A Philosophical Essay on Probabilities”.
ISBN 1602063281, translated from the French 6th ed. (1840).
Laud P, Ibrahim J (1995). “Predictive Model Selection.” Journal of the Royal Statistical
Society, B 57, 247–262.
Lewis S, Raftery A (1997). “Estimating Bayes’ Factors via Posterior Simulation with the
Laplace-Metropolis Estimator.” Journal of the American Statistical Association, 92, 648–
655.
Meng X (1994). “Posterior Predictive P-Values.” Annals of Statistics, 22, 1142–1160.
Ntzoufras I (2009). Bayesian Modeling Using WinBUGS. John Wiley & Sons, West Sussex,
England.
R Core Team (2014). R: A Language and Environment for Statistical Computing. R Foun-
dation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org.
Raiffa H, Schlaifer R (1961). Applied Statistical Decision Theory. Division of Research,

Graduate School of Business Administration, Harvard University.
Ramsey F (1926). “Truth and Probability.” In R Braithwaite (ed.), The Foundations of

Mathematics and other Logical Essays, pp. 156–198. Harcourt, Brace and Company, New
York.
Robert C (2007). The Bayesian Choice. 2nd edition. Springer, Paris, France.
Rossi P, Allenby G, McCulloch R (2005). Bayesian Statistics and Marketing. John Wiley &
Sons, West Sussex, England.
Savage L (1954). The Foundations of Statistics. John Wiley & Sons, West Sussex, England.
Spiegelhalter D, Best N, Carlin B (1998). “Bayesian Deviance, the Effective Number of

Parameters, and the Comparison of Arbitrarily Complex Models.” Research Report, 98-
009.
Spiegelhalter D, Best N, Carlin B, van der Linde A (2002). “Bayesian Measures of Model
Complexity and Fit (with Discussion).” Journal of the Royal Statistical Society, B 64,
583–639.
Spiegelhalter D, Thomas A, Best N, Lunn D (2003). WinBUGS User Manual, Version 1.4.
MRC Biostatistics Unit, Institute of Public Health and Department of Epidemiology and
Public Health, Imperial College School of Medicine, UK.
Statisticat LLC (2015). LaplacesDemon: Complete Environment for Bayesian Inference. R

package version 15.03.19, URL http://www.bayesian-inference.com/software.
Stigler S (1983). “Who Discovered Bayes’s Theorem?” The American Statistician, 37(4),
290–296.
Affiliation:
Statisticat, LLC
Farmington, CT
E-mail: is defunct
URL: https://web.archive.org/web/20141224051720/http://www.bayesian-inference.
com/index

Hipnotic Bayestas

Uploaded by

Copyright:

Available Formats

Hipnotic Bayestas

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Hipnotic Bayestas

Uploaded by

Copyright:

Available Formats

Bayesian Inference

Keywords: Bayesian, LaplacesDemon, LaplacesDemonCpp, R.

1.1. Bayes’ Theorem, Example 1

1.2. Bayes’ Theorem, Example 2

• Pr(A1 |B) = Pr(Hell|Consort)

• Pr(A2 |B) = Pr(Heaven|Consort)

• 6 people consorted out of 9 who went to Hell.

• 5 people consorted out of 7 who went to Heaven.

• 75% of the population goes to Hell.

• 25% of the population goes to Heaven.

• Pr(Consort|Hell) = 6/9 = 0.666

• Pr(Consort|Heaven) = 5/7 = 0.714

2. Model-Based Bayesian Inference

p(Θ|y) ∝ p(y|Θ)p(Θ) (2)

3. Components of Bayesian Inference

4.1. Informative Priors

4.2. Weakly Informative Priors

where θ is normally-distributed according to a mean of 0 and a variance of 10,000, which is

4.3. Least Informative Priors

where θ is uniformly-distributed from negative infinity to positive infinity. Although this

4.4. Uninformative Priors

4.5. Proper and Improper Priors

As noted previously, an unbounded uniform prior distribution is an improper prior distribution

The simplest hierarchical Bayes model takes the form

p(Θ, Φ|y) ∝ p(y|Θ)p(Θ|Φ)p(Φ)

When the half-Cauchy is unavailable, a uniform distribution is often placed on σ in hierarchical

Conjugacy is mathematically convenient in that the posterior distribution follows a known

7.1. Terminology: From Inverse Probability to Bayesian Probability

7.2. The Likelihood Principle

Frequentist “procedures typically violate the likelihood principle, since long-run

There is “a general fact about the mechanism of parametric Bayesian inference

“Although the likelihood principle is implicit in Bayesian statistics, it was devel-

“The likelihood principle, by itself, is not sufficient to build a method of inference

“Adherence to the likelihood principle means that inferences are ‘conditional’ on

is worth contrasting to any sampling-based approach to inference. In the sam-

7.3. Likelihood Function of a Parameterized Model

L (θ|y) = p(y|θ) = f (y|θ)

where y is the observed outcome of an experiment, and the likelihood (L ) of θ given y is

The dependent variable y, indexed by i = 1, ..., n, is stochastic, and normally-distributed

p(β, σ 2 |y) = p(y|β, σ 2 )p(β1 )p(β2 )p(β3 )p(σ 2 )

log[p(β, σ 2 |y)] = log[p(y|β, σ 2 )] + log[p(β1 )] + log[p(β2 )] + log[p(β3 )] + log[p(σ 2 )]

The posterior predictive distribution is easy to estimate

10. Bayes Factors

where p(y|M1 ) is the marginal likelihood of the data in model 1.

(Posterior model odds) = (Bayes factor) x (prior model odds)

p(y|m) = (2π)dm /2 |Σm |1/2 p(y|Θm , m)p(Θm |m)

2. Deviance Information Criterion (DIC)

3. Widely Applicable Information Criterion (WAIC)

11. Model Fit

12. Posterior Predictive Checks

12.1. Bayesian p-values

[yi − E(yi )]2

12.3. Conditional Predictive Ordinate

12.4. Predictive Concordance

over t = 1, . . . , T posterior samples. The calibration number, SL , is the standard deviation of

13. Advantages Of Bayesian Inference Over Frequentist Inference

• Bayesian inference allows informative priors so that prior knowledge or results of a

• Bayesian inference estimates p(hypothesis|data). In contrast, frequentist inference esti-