Transcript Lecture 10
Bayesian Model Selection in
Factorial Designs
Seminal work is by Box and Meyer
Intuitive formulation and analytical
approach, but the devil is in the details!
Look at simplifying assumptions as we
step through Box and Meyer’s approach
One of the hottest areas in statistics for
several years
Bayesian Model Selection in
Factorial Designs
There are 2k-p-1=n-1 possible (fractional)
factorial models, denoted as a set {Ml}.
To simplify later calculations, we usually
assume that the only active effects are
main effects, two-way effects or three-way
effects
– This assumption is already in place for lowresolution fractional factorials
Bayesian Model Selection in
Factorial Designs
Each Ml denotes a set of active effects
(both main effects and interactions) in a
hierarchical model.
We will use Xik=1 for the high level of
effect k and Xik=-1 for the low level of
effect k.
Bayesian Model Selection in
Factorial Designs
We will assume that the response
variables have a linear model with normal
errors given model M
Xi and b are model-specific, but we will
use a saturated model in what follows
Bayesian Model Selection in
Factorial Designs
The likelihood for the data given the
parameters has the following form
2n
æ 1
2ö
L(b, s ,Y ) = Õ
exp ç - 2 (Yi - Xi¢b ) ÷
2
è 2s
ø
2ps
i=1
æ 1
ö
1
¢
=
exp ç - 2 (Y - X b ) (Y - X b ) ÷
n
2 2
è
ø
2
s
2ps
(
)
1
Bayesian Paradigm
Unlike in classical inference, we assume
the parameters, Q, are random variables
that have a prior distribution, fQ(q), rather
than being fixed unknown constants.
In classical inference, we estimate q by
maximizing the likelihood L(q|y)
Bayesian Paradigm
Estimation using the Bayesian approach
relies on updating our prior distribution for
Q after collecting our data y. The posterior
density, by an application of Bayes rule, is
proportional to the product of the familiar
data density and the prior density:
fQ|Y (q | y ) µ f X |Q (x | q ) fQ (q )
Bayesian Paradigm
The Bayes estimate of Q minimizes Bayes
risk—the expected value (with respect to
the prior) of loss function L(q).
Under squared error loss, the Bayes
estimate is the mean of the posterior
distribution:
qˆ ( y ) = EQ|Y Q
Bayesian Model Selection in
Factorial Designs
The Bayesian prior for models is quite
straightforward. The prior probability that r
effects are active in the model, given each
is active with prior probability p, is
L(p ) = C1p (1- p )
r
n-1-r
= C1 (1- p )
n-1
r
æ p ö
ç
÷
è 1- p ø
Bayesian Model Selection in
Factorial Designs
Since we’re using a Bayesian approach,
we need priors for b and s as well
b0 ~ N (0, s e ), e = 10
2
b j ~ N (0, g s
2
2
)
s ~ g(s ), g(s ) µ s
-a
-6
Bayesian Model Selection in
Factorial Designs
For non-orthogonal designs, it’s common
to use Zellner’s g-prior for b:
(
b j ~ N 0, g s ( X' X )
2
2
-1
)
Note that we did not assign priors to g or
p
Bayesian Model Selection in
Factorial Designs
We can combine f(b,s,M) and f(Y|b,s,M)
to obtain the full likelihood L(b,s,M,Y)
æ1ö æ 1 ö
÷
ç ÷ ç
è g ø è 2ps 2 ø
æ 1
ö
exp ç - 2 Q(b )÷
è 2s
ø
r
æ p ö
L(b, s , M,Y ) = C ç
÷
è 1- p ø
n-1
2n-1+a
2
´
Bayesian Model Selection in
Factorial Designs
Bayesian Model Selection in
Factorial Designs
Our goal is to derive the posterior
distribution of M given Y, which first
requires integrating out b and s.
L(M | Y ) µ L(M,Y ) =
¥
ò ò L(b, s , M,Y )d b ds =
0 Rn
r
æ p ö
Cç
÷
è 1- p ø
( X ¢X + G)
n-1
æ1ö
ç ÷
èg ø
(
-1 2
Q ( X ¢X + G) X ¢Y
-1
)
(n-1+a) 2
Bayesian Model Selection in
Factorial Designs
The first term is a penalty for model
complexity (smaller is better)
The second term is a measure of model fit
(smaller is better)
Bayesian Model Selection in
Factorial Designs
p and g are still present. We will fix p; the
method is robust to the choice of p
g is selected to minimize the probability of
no active factors
Bayesian Model Selection in
Factorial Designs
With L(M|Y) in hand, we can actually
evaluate the P(Mi|Y) for all Mi for any prior
choice of p, provided the number of Mi is
not burdensome
This is in part why we assume eligible Mi
only include lower order effects.
Bayesian Model Selection in
Factorial Designs
Greedy search or MCMC algorithms are
used to select models when they cannot
be itemized
Selection criteria include Bayes Factor,
Schwarz criterion, Bayesian Information
Criterion
Refer to R package BMA and bic.glm for
fitting more general models.
Bayesian Model Selection in
Factorial Designs
For each effect, we sum the probabilities
for all Mi that contain that effect and obtain
a marginal posterior probability for that
effect.
These marginal probabilities are relatively
robust to the choice of p.
Case Study
Violin data* (24 factorial design with n=11
replications)
Response: Decibels
Factors
–
–
–
–
A: Pressure (Low/High)
B: Placement (Near/Far)
C: Angle (Low/High)
D: Speed (Low/High)
*Carla Padgett, STAT 706 taught by Don Edwards
Case Study
Fractional Factorial
Design:
• A, B, and D significant
• AB marginal
Bayesian Model
Selection:
• A, B, D, AB, AD,
BD significant
• All others
negligible
*Carla Padgett, STAT 706 taught by Don Edwards