Comparing sampling models by Bayes factors

In the last posting, I illustrated fitting a t(4) sampling model to some baseball data where I suspected there was an outlier. To compare this model to other sampling models, we can use Bayes factors.

Here is a t sampling model with a convenient choice of prior.

1.is a random sample from
2. and are independent, with distributed and distributed

I define a R function tsampling.R that computes the logarithm of the posterior. Note that I am careful to include all of the normalizing constants, since I am primarily interested in computing the marginal density of .

tsampling=function(theta,datapar)
{
mu=theta[1]; logsigma=theta[2]; sigma=exp(logsigma)

y=datapar$y; df=datapar$df
mu.mean=datapar$mu.mean; mu.sd=datapar$mu.sd
lsigma.mean=datapar$lsigma.mean; lsigma.sd=datapar$lsigma.sd

loglike=sum(dt((y-mu)/sigma,df,log=TRUE)-log(sigma))
logprior=dnorm(mu,mu.mean,mu.sd,log=TRUE)+
dnorm(logsigma,lsigma.mean,lsigma.sd,log=TRUE)

return(loglike+logprior)
}

We use this function together with the function laplace to compute the log marginal density for two models.

Model 1 -- t(4) sampling, is normal(90, 20), is normal(1, 1).

Model 2 -- t(30) sampling, is normal(90, 20), is normal(1, 1).

Note that I'm using the same prior for both models -- the only difference is the choice of degrees of freedom in the sampling density.

dataprior1=list(y=y, df=4, mu.mean=90, mu.sd=20,
lsigma.mean=1, lsigma.sd=1)
log.m1=laplace(tsampling,c(80, 3), dataprior1)$int

dataprior2=list(y=y, df=30, mu.mean=90, mu.sd=20,
lsigma.mean=1, lsigma.sd=1)
log.m2=laplace(tsampling,c(80, 3), dataprior2)$int

BF.12=exp(log.m1-log.m2)
BF.12
[1] 14.8463

We see that there is substantial support for the t(4) model over the "close to normal" t(30) model.

Robust Modeling using Gibbs Sampling

To illustrate the use of Gibbs sampling for robust modeling, here are the batting statistics for the "regular" players on the San Francisco Giants for the 2007 baseball season:

Pos Player              Ag   G   AB    R    H   2B 3B  HR  RBI  BB  SO   BA    OBP   SLG  SB  CS  GDP HBP  SH  SF IBB  OPS+
---+-------------------+--+----+----+----+----+---+--+---+----+---+----+-----+-----+-----+---+---+---+---+---+---+---+----+
C   Bengie Molina       32  134  497   38  137  19  1  19   81  15   53  .276  .298  .433   0   0  13   2   1   2   2   86
1B *Ryan Klesko         36  116  362   51   94  27  3   6   44  46   68  .260  .344  .401   5   1  14   1   1   1   2   92
2B #Ray Durham          35  138  464   56  101  21  2  11   71  53   75  .218  .295  .343  10   2  18   2   0   9   6   65
3B  Pedro Feliz         32  150  557   61  141  28  2  20   72  29   70  .253  .290  .418   2   2  15   1   0   3   2   81
SS #Omar Vizquel        40  145  513   54  126  18  3   4   51  44   48  .246  .305  .316  14   6  14   1  14   3   6   62
LF *Barry Bonds         42  126  340   75   94  14  0  28   66 132   54  .276  .480  .565   5   0  13   3   0   2  43  170
CF *Dave Roberts        35  114  396   61  103  17  9   2   23  42   66  .260  .331  .364  31   5   4   0   4   0   1   80
RF #Randy Winn          33  155  593   73  178  42  1  14   65  44   85  .300  .353  .445  15   3  12   7   4   5   3  105
    Rich Aurilia        35   99  329   40   83  19  2   5   33  22   45  .252  .304  .368   0   0   8   4   0   3   1   73
    Kevin Frandsen      25  109  264   26   71  12  1   5   31  21   24  .269  .331  .379   4   3  17   5   3   3   3   84
   *Fred Lewis          26   58  157   34   45   6  2   3   19  19   32  .287  .374  .408   5   1   4   3   1   0   0  103
   #Dan Ortmeier        26   62  157   20   45   7  4   6   16   7   41  .287  .317  .497   2   1   2   1   0   2   1  107
    Rajai Davis         26   51  142   26   40   9  1   1    7  14   25  .282  .363  .380  17   4   0   4   2   0   1   93
   *Nate Schierholtz    23   39  112    9   34   5  3   0   10   2   19  .304  .316  .402   3   1   0   1   0   2   0   85
   *Mark Sweeney        37   76   90   18   23   8  0   2   10  13   18  .256  .368  .411   2   0   0   3   1   0   0  102

We'll focus on the last batting measure OPS that is a summary of a player's batting effectiveness.

We read the OPS values for the 15 players into R into the vector y.

> y
[1] 86 92 65 81 62 170 80 105 73 84 103 107 93 85 102

We assume that the observations y1, ..., y15 are iid from a t distribution with location , scale and 4 degrees of freedom. We place the usual noninformative prior on .

We implement 10,000 iterations of Gibbs sampling by use of the function robustt.R in the LearnBayes package. This function is easy to use -- we just input the data vector y, the degrees of freedom, and the number of iterations.

fit=robustt(y,4,10000)

The object fit is a list with components mu, sigma2, and lam -- mu is a vector of simulated draws of , sigma2 is a vector of simulated draws of , and lam is a matrix of simulated draws of the scale parameters . (The are helpful for identifying outliers in the data -- here the outlier is pretty obvious.)

Below I have graphed the data as solid dots and placed a density estimate of the posterior of on top. We see that the estimate of appears to ignore the one outlier (Barry Bonds) in the data.

plot(y,0*y,cex=1.5,pch=19,ylim=c(0,.1),ylab="DENSITY",xlab="MU")
lines(density(fit$mu),lwd=3,col="red")

By the way, we have assumed that robust modeling was suitable for this dataset since I knew that the Giants had one unusually good hitter on their team. Can we formally show that robust modeling the t(4) distribution is better than normal modeling for these data?

Sure -- we can define two models (with the choice of proper prior distributions) and compare the models by use of a Bayes factor. I'll illustrate this in my next posting.

Gibbs Sampling for Censored Regression

Gibbs sampling is very convenient for many "missing data" problems. To illustrate this situation, suppose we have the simple regression model

,

where the errors are iid . The problem is that some of the response variables are right-censored and we actually observe , where is a known censoring value. We know which observations are uncensored () and which observations are censored ().

We illustrate this situation by use of a picture inspired by a similar one in Martin Tanner's book. We show a scatterplot of the covariate against the observed . The points highlighted in red correspond to censored observations -- the actual observations (the ) exceed the censored values, which we indicate by arrows.


To apply Gibbs sampling to this situation, we imagine a complete data set where all of the 's are known. The unknowns in this problem are the regression coefficients , the error variance , and the 's corresponding to the censored observations.

The joint posterior of all unobservables (assuming the usual vague prior for regression) has the form

where is equal to 1 if the observation is not censored, and if the observation is censored at .

With the introduction of the complete data set, Gibbs sampling is straightforward.

Suppose one has initial estimates at the regression coefficients and the error variance. Then

(1) One simulates from the distribution of the missing data (the for the censored observations) given the parameters. Specifically is simulated from a normal() distribution censored below by .

(2) Given the complete data, one simulates values of the parameters using the usual approach for a normal linear regression model.

To do this on R, I have a couple of useful tools. The function rtruncated.R will simulate draws from an arbitrary truncated distribution

rtruncated=function(n,lo,hi,pf,qf,...)
qf(pf(lo,...)+runif(n)*(pf(hi,...)-pf(lo,...)),...)

For example, if one wishes to simulate 20 draws of a normal(mean = 10, sd = 2) distribution that is truncated below by 4, one writes

rtruncated(20, 4, Inf, pnorm, qnorm, 10, 2)

Also the function blinreg.R in the LearnBayes package will simulate draws of a regression coefficient and the error variance for a normal linear model with a noninformative prior.

Latex in my blog?

I didn't know if it was possible to add latex to my postings. I asked John Shonder who is currently working on solutions on my book and has some latex in his WordPress blog. John referred me to a page

http://wolverinex02.googlepages.com/emoticonsforblogger2

that describes a simple procedure for typing in latex in one's postings.

Anyway, I tried it out on the "Gibbs sampling for hierarchical models" posting and it works!

A Comparison of Two MCMC Methods

In the last posting, I illustrated a Gibbs sampling algorithm for simulating from a normal/normal hierarchical model. This method was based on successive substitution sampling from the conditional posterior densities of theta1, ..., thetak, mu, and tau2. A second sampling method is outlined in Chapter 7 of BCUR. In this method that we'll call "Exact MCMC" one integrates out the first-stage parameters theta1, ..., thetak and uses a random-walk Metropolis algorithm to simulate from the marginal posterior of (mu, log tau).

We'll use a few pictures to compare these two methods, specifically in learning about the variance hyperparameter tau. Here's the exact MCMC method:

1. We write a function defining the log posterior of mu and log tau.

2. We simulate from (mu, log tau) using the function rwmetrop.R in the LearnBayes package. We choose a suitable proposal function from output from the laplace.R function. The acceptance rate of this random walk algorithm is 20%. We sample for 10,000 iterations, saving the draws of log tau in a vector.

We then run the Gibbs sampler for 10,000 iterations, also saving the draws of log tau.

DENSITY PLOTS. We first compare density estimates of the simulated samples of log tau using the two methods. The two estimates seem pretty similar. The density estimate for the Gibbs sampling draws looks smoother, but that doesn't mean this sample is a better estimate of the posterior of log tau.



TRACE PLOTS. We compare trace plots of the two sets of simulated draws. They look pretty similar, but there are differences in closer inspection. The draws from the Gibbs sampling run look more irregular, or perhaps they have more of a "snake-like" appearance.



AUTOCORRELATION PLOTS. A autocorrelation plot is a graph of the autocorrelations

corr (theta(j), theta(j-g))

graphed against the lag g. For most MCMC runs, the values of the stream of simulated draws will show positive autocorrelation, but hopefully the autocorrelation values decrease quickly as the lag increases. It is pretty obvious that the lag autocorrelation values drop off slower for method 2 (Gibbs sampling); in contrast, the lag autocorrelations decrease to zero for method 1 (Exact MCMC) for values of the lag from 1 to 15.


The moral of the story is that the exact MCMC method shows better mixing than Gibbs sampling in this example. This means that, for a given sample size (here 10,000), one will have more accurate estimates at the posterior of tau using the exact MCMC method.

Gibbs Sampling for Hierarchical Models

Gibbs sampling is an attractive "automatic" method of setting up a MCMC algorithm for many classes of models. Here we illustrate using R to write a Gibbs sampling algorithm for the normal/normal exchangeable model.

We write the model in three stages as follows.

1. The observations y1, ..., yk are independent where yj is N(thetaj, sigmaj^2), where we write N(mu, sigma2) to denote the normal density with mean mu and variance sigma2. We assume the sampling variances sigma1^2, ..., sigmak^2 are known.

2. The means theta1,..., thetak are a random sample from a N(mu, tau2) population. (tau2 is the variance of the population).

3. The hyperparameters (mu, tau) are assumed to have a uniform distribution. This implies that the parameters (mu, tau2) have a prior proportional to (tau2)^(-1/2).

To write a Gibbs sampling algorithm, we write down the joint posterior of all parameters (theta1, ..., thetak, mu, tau2):




From this expression, one can see

1. The posterior of thetaj conditional on all remaining parameters is normal, where the mean and variance are given by the usual normal density/normal prior updating formula.

2. The hyperparameter mu has a normal posterior with mean theta_bar (the sample mean of the thetaj) and variance tau2/k.

3. The hyperparameter tau2 has an inverse gamma posterior with shape (k-1)/2 and rate 1/2 sum(thetaj - mu)^2.

Given that all the conditional posteriors have convenient functional forms, we write a R function to implement the Gibbs sampling. The only inputs are the data matrix (columns containing yj and sigmaj^2) and the number of iterations m.

I'll display the function normnormexch.gibbs.R with notes.

normnormexch.gibbs=function(data,m)
{
y=data[,1]; k=length(y); sigma2=data[,2] # HERE I READ IN THE DATA

THETA=array(0,c(m,k)); MU=rep(0,m); TAU2=rep(0,m) # SET UP STORAGE

mu=mean(y); tau2=median(sigma2) # INITIAL ESTIMATES OF MU AND TAU2

for (j in 1:m) # HERE'S THE GIBBS SAMPLING
{
p.means=(y/sigma2+mu/tau2)/(1/sigma2+1/tau2) # CONDITIONAL POSTERIORS
p.sds=sqrt(1/(1/sigma2+1/tau2)) # OF THETA1,...THETAK
theta=rnorm(k,mean=p.means,sd=p.sds)

mu=rnorm(1,mean=mean(theta),sd=sqrt(tau2/k)) # CONDITIONAL POSTERIOR OF MU

tau2=rigamma(1,(k-1)/2,sum((theta-mu)^2)/2) # CONDITIONAL POSTERIOR OF TAU2

THETA[j,]=theta; MU[j]=mu; TAU2[j]=tau2 # STORE SIMULATED DRAWS
}
return(list(theta=THETA,mu=MU,tau2=TAU2)) # RETURN A LIST WITH SAMPLES
}

Here is an illustration of this algorithm for the SAT example from Gelman et al.

y=c(28,8,-3,7,-1,1,18,12)
sigma=c(15,10,16,11,9,11,10,18)
data=cbind(y,sigma^2)
fit=normnormexch.gibbs(data,1000)

In the Chapter 7 exercise that used this example, a different sampling algorithm was used to simulate from the joint posterior of (theta, mu, tau2) -- it was a direct sampling algorithm based on the decomposition

[theta, mu, tau2] = [mu, tau2] [theta | mu, tau2]

In a later posting, I'll compare the two sampling algorithms.

Bayesian model selection

Here we illustrate one advantage of Bayesian regression modeling. By the use of an informative prior, it is straightforward to implement regression model selection.

Arnold Zellner introduced an attractive method of implementing prior information into a regression model. He assumes [beta | sigma^2] is normal with mean beta0 and variance-covariance matrix of the form

V = c sigma^2 (X'X)^(-1)

and then takes [sigma^2] distributed according to the noninformative prior proportional to 1/sigma^2. This is called Zellner's G-prior.

One nice thing about this prior is that it requires only two prior inputs from the user: (1) a choice of the prior mean beta0, and (2) c that can be interpreted as the amount of information in the prior relative to the sample.

We can use Zellner's G-prior to implement model selection in a regression analysis. Suppose we have p predictors of interest -- there are 2^p possible models corresponding to the inclusion or exclusion of each predictor in the model.

A G-prior is placed on the full model that contains all parameters. We assume that beta0 is the zero vector and choose c to be a large value reflecting vague prior beliefs. The prior on (beta, sigma^2) for this full model is

N(beta; beta0, c sigma^2 (X'X)^(-1)) (1/sigma^2)

Then for any submodel defined by a reduced design matrix X1, we take the prior on (beta, sigma^2) to be

N(beta; beta0, c sigma^2 (X1'X1)^(-1)) (1/sigma^2)

Then we can compare models by the computation of associated predictive probabilities.

To illustrate this methodology, we consider an interesting dataset on the behavior of puffins from Devore and Peck's text. One is interesting in understanding the nesting frequency behavior of these birds (the y variable) on the basis of four covariates: x1, the grass cover, x2, the mean soil depth, x3, the angle of slope, and x4, the distance from cliff edge.

We first write a function that computes the log posterior of (beta, log(sigma)) for a regression model with normal sampling and a Zellner G prior with beta0 = 0 and a given value of c.

regpost.mod=function(theta,stuff)
{
y=stuff$y; X=stuff$X; c=stuff$c
beta=theta[-length(theta)]; sigma=exp(theta[length(theta)])
if (length(beta)>1)
loglike=sum(dnorm(y,mean=X%*%as.vector(beta),sd=sigma,log=TRUE))
else
loglike=sum(dnorm(y,mean=X*beta,sd=sigma,log=TRUE))
logprior=dmnorm(beta,mean=0*beta,varcov=c*sigma^2*solve(t(X)%*%X),log=TRUE)
return(loglike+logprior)
}

We read in the puffin data and define the design matrix X for the full model.

puffin=read.table("puffin.txt",header=T)
X=cbind(1,puffin$x1,puffin$x2,puffin$x3,puffin$x4)

S, the input for the log posterior function, is a list with components y, X, and c.

S=list(y=puffin$y, X=X, c=100)

Since there are 4 covariates, there are 2^4 = 16 possible models. We define a logical matrix GAM of dimension 16 x 5 that describes the inclusion of exclusion of each covariate in the model. (The first column is TRUE since we want to include the constant term in each regression model.)

GAM=array(T,c(16,5))
TF=c(T,F); k=0
for (i1 in 1:2) {for (i2 in 1:2) {for (i3 in 1:2) {for (i4 in 1:2){
k=k+1; GAM[k,]=cbind(T,TF[i1],TF[i2],TF[i3],TF[i4])}}}}

For each model, we use the laplace function (in the LearnBayes package) to compute the marginal likelihood. The inputs to laplace are the function regpost.mod defining our model, an intelligent guess at the model (given by a least-squares fit), and the list S that contains y, X, and c.

gof=rep(0,16)
for (j in 1:16)
{
S$X=X[,GAM[j,]]
theta=c(lm(S$y~0+S$X)$coef,0)
gof[j]=laplace(regpost.mod,theta,S)$int
}

We display each model and the associated marginal likelihood values (on the log scale).

data.frame(GAM,gof)

X1 X2 X3 X4 X5 gof
1 TRUE TRUE TRUE TRUE TRUE -104.1850
2 TRUE TRUE TRUE TRUE FALSE -115.4042
3 TRUE TRUE TRUE FALSE TRUE -102.3523
4 TRUE TRUE TRUE FALSE FALSE -136.3972
5 TRUE TRUE FALSE TRUE TRUE -105.0931
6 TRUE TRUE FALSE TRUE FALSE -113.1782
7 TRUE TRUE FALSE FALSE TRUE -105.5690
8 TRUE TRUE FALSE FALSE FALSE -134.0486
9 TRUE FALSE TRUE TRUE TRUE -101.8833
10 TRUE FALSE TRUE TRUE FALSE -114.9573
11 TRUE FALSE TRUE FALSE TRUE -100.3735
12 TRUE FALSE TRUE FALSE FALSE -134.5129
13 TRUE FALSE FALSE TRUE TRUE -102.8117
14 TRUE FALSE FALSE TRUE FALSE -112.6721
15 TRUE FALSE FALSE FALSE TRUE -103.2963
16 TRUE FALSE FALSE FALSE FALSE -132.1824

I highlight the model (inclusion of covariates X3 and X5) that has the largest value of the log marginal likelihood. This tells us that the best model for understanding nesting behavior is the one that includes mean soil depth (X3) and the distance from cliff edge (X5) . One can also compare different models by the use of Bayes factors.

****************************************************************************
For students who have to do their own model selection, I've written a simple function

bayes.model.selection.R

that will give these log marginal likelihood values for all regression models. You have to download this function from bayes.bgsu.edu/m648 and have LearnBayes 1.11 installed on your machine.

Here's an example of using this function. I first load in the puffin dataset and define the response vector y, the covariate matrix X, and the prior parameter c.

puffin=read.table("puffin.txt",header=T)
X=cbind(1,puffin$x1,puffin$x2,puffin$x3,puffin$x4)
y=puffin$y
c=100

Then I just run the function bayes.model.selection with y, X, and c as inputs.

bayes.model.selection(y,X,c)

The output will be the data frame shown above.

Bayesian regression

To introduce Bayesian regression modeling, we consider a dataset from De Veaux, Velleman and Bock which collected physical measurements from a sample of 250 males. One is interested in predicting a person's body fat from his height, waist, and chest measurements.

The file Body_fat.txt contains the data that we read in R.

data=read.table("Body_fat.txt",sep="\t",header=TRUE)
names(data)
[1] "Pct.BF" "Height" "Waist" "Chest"
attach(data)

Suppose we wish to fit the regression model

Pct.BF ~ Height + Waist + Chest

The standard least-squares fit is done using the lm command in R.

fit=lm(Pct.BF~Height+Waist+Chest)

Here is a portion of the summary of the fit.

summary(fit)

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 2.06539 7.80232 0.265 0.79145
Height -0.56083 0.10940 -5.126 5.98e-07 ***
Waist 2.19976 0.16755 13.129 < 2e-16 ***
Chest -0.23376 0.08324 -2.808 0.00538 **
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 4.399 on 246 degrees of freedom
Multiple R-Squared: 0.7221, Adjusted R-squared: 0.7187
F-statistic: 213.1 on 3 and 246 DF, p-value: < 2.2e-16

Now let's consider a Bayesian fit of this model. Suppose we place the usual noninformative prior on the regression vector beta and the error variance sigma^2.

g(beta, sigma^2) = 1/sigma^2.

Then there is a simple direct method (outlined in BCUR) of simulating from the posterior distribution of (beta, sigma^2).

1. We first create the design matrix X:

X=cbind(1,Height,Waist,Chest)

The response is contained in the vector Pct.BF.

2. To simulate 5000 draws from the posterior of (beta, sigma), we use the function blinreg in the LearnBayes package.

fit=blinreg(Pct.BF, X, 5000)

The output fit is a list with two components: beta is a matrix of simulated draws of beta where each column corresponds to a sample from component of beta, and sigma is a vector of draws from the marginal posterior of sigma.

We can summarize the simulated draws of beta by computing posterior means and standard deviations.

apply(fit$beta,2,mean)
X XHeight XWaist XChest
1.8748122 -0.5579671 2.2031474 -0.2350661

apply(fit$beta,2,sd)
X XHeight XWaist XChest
7.84069390 0.11050071 0.16839919 0.08286758

Here is a graph of density estimates of simulated draws from the four regression parameters.

par(mfrow=c(2,2))
for (j in 1:4) plot(density(fit$beta[,j]),main=paste("BETA ",j),
lwd=3, col="red", xlab="PAR")

What's so special about Bayesian regression if we are essentially replicating the frequentist regression fit?

We'll talk about the advantages of Bayesian regression in the next blog posting.

Looking for True Streakiness

There is a lot of interest in streaky behavior in sports. One observes players or teams that appear streaky with the implicit conclusion that this says something about the character of the athlete.

Eric Byrnes had 412 opportunities to hit during the 2005 baseball season. Here is his sequence of hits (successes) and outs (failures) during the season.

[1] 0 0 0 1 0 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 1 0 0 1 0 0 1 0 0 1 0 1 0 0 0 1 0
[38] 0 1 1 1 0 0 0 1 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 1 1 0 0 0 0 0 1 0
[75] 1 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 1 0 0 0
[112] 0 0 1 0 1 1 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 1 0 0 0 0 0 0 1 0 1 0 0 0 0 0
[149] 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 1 1 0 0 0 0 1 0
[186] 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 1 1 0 0 0 0 0 1 0 1 0 0 1 0 1 0 0 1 0 0 0 0
[223] 0 0 0 0 0 0 0 0 0 1 0 1 1 0 1 0 0 1 1 0 0 0 1 1 0 1 1 0 1 1 0 1 1 0 1 0 0
[260] 0 0 0 0 1 0 1 0 0 0 0 1 1 1 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0
[297] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 1 0 0 1 0 0 0 0 1 0
[334] 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0
[371] 0 0 1 1 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 1 1 0 0 0 0 1 0 1 0 1 1 1 0 0
[408] 0 0 0 0 0

One way of seeing the streaky behavior in this sequence is by a moving average graph where one plots the success rate (batting average) for windows of 40 at-bats. I wrote a short program mavg.R to compute the moving averages. The following R code plots the moving averages and plots a lowess smooth on top to help see the pattern.

MAVG=mavg(byrne$x,k=40)
plot(MAVG,type="l",lwd=2,col="red",xlab="GAME",ylab="AVG",
main="ERIC BYRNES")

We some interesting patterns.. It seemed that Byrnes had a cold spell in the first part of the season, followed by a hot period, and then a very cold period.

The interesting question is: is this streaky pattern "real" or is it just a byproduct of bernoulli chance variation?

We answer this question by means of a Bayes factor. Suppose we partition Byrnes' 412 at-bats into groups of 20 at-bats. We observe counts y1, ..., yn, where yi is the number of hits in the ith group. Suppose yi is binomial(20, pi) where pi is the probability of a hit in the ith period.

We define two hypotheses:

H (not streaky) the probabilities across periods are equal, p1 = ... = pn = p

A (streaky) the probabilities across periods vary according to a beta distribution with mean eta and precision K. This model is indexed by the parameter K.

The functions bfexch and laplace in the LearnBayes package can be used to compute a Bayes factor in support of A over H. Here is how we do it.

1. The raw data is in the matrix BRYNE -- the first column contains the data (0's and 1's) and the second column contains the attempts (column of 1's). We regroup the data into periods of 20 at-bats using the regroup function.

regroup(BRYNE, 20)

2. The following R function laplace.int will compute the log, base10 of the Bayes factor in support of streakiness for a fixed value of log(K).

laplace.int=function(logK,data=data1)
log10(exp(laplace(bfexch,0,list(data=data,K=exp(logK)))$int))

To illustrate, suppose we want to compute the log10 Bayes factor for our data for logK = 3:

> laplace.int(3,regroup(BRYNE,20))
[,1]
[1,] 1.386111

This indicates support for streakiness -- the log Bayes factor is 1.38 which means that A is over 10 times more likely than H.

3. Generally we'd like to compute the log10 Bayes factor for a sequence of values of log K. I first write a simple function that does this:

s.laplace.int=function(logK,data)
list(x=logK,y=sapply(logK,laplace.int,data))

and then I use this function to compute the Bayes factor for values of log K from 2 to 6 in steps of 0.2. I use the plot command to graph these values. I draw a line at the value log10 BF = 0 -- this corresponds to the case where neither model is supported.

plot(s.laplace.int(seq(2,6,by=.2),regroup(BRYNE,20)),type="l",
xlab="LOG K", ylab="LOG 10 BAYES FACTOR", lwd=3, col="red", ylim=c(-3,2))
lines(c(1,7),c(0,0),lwd=3,col="blue")
title(main="ERIC BYRNES")

What we see that, for a range of values of K, the Bayes factor favors the model A by a factor of 10 or more.

Actually we only looked at Eric Byrnes since he exhibited unusually streaky behavior during this 2005 season. What if we look at other players? Here are the Bayes factors graphs for the hitting data for two other players Chase Utley and Damian Miller (we are grouping the data in the same way).


Here for both players, note that the log10 Bayes factors are entirely negative for the range of K values. For both players, there is support for the non-streaky model H. One distinctive features of Bayes factors is that they can provide support for the null or the alternative hypothesis.

Test of Independence in a 2 x 2 Table

Consider data collected in a study described in Dorn (1954) to assess the relationship between smoking and lung cancer. In this study, a sample of 86 lung-cancer patients and a sample of 86 controls were questioned about their smoking habits. The two groups were chosen to represent random samples from a sub-population of lung-cancer patients and an otherwise similar population of cancer-free individuals.

Here is the 2 x 2 table of responses:

Cancer Control
Smokers 83 72
Non-smokers 3 14

Let pL and pC denote the proportions of lung-cancer patients and controls who smoke. We wish to test H: pL = pC against the alternative A: pL <> pC.

To construct a Bayesian test, we define a suitable model for H and for A, and then compute the Bayes factor in support of the alternative A.

1. To describe these models, first we transform the proportions to the logits

LogitL = log(pL/(1-pL)), Logit C = log(pC/(1-pC))

2. We then define two parameters theta1, theta2, that are equal to the difference and sum of the logits.

theta1 = LogitL - LogitC, theta2 = LogitL + LogitC.

theta1 is the log odds ratio, a popular measure of association in a 2 x 2 table. Under the hypothesis of independence H, theta1 = 0.

3. Consider the following prior on theta1 and theta2. We assume they are independent where

theta1 is N(0, tau1), theta2 is N(0, tau2).

4. Under H (independence), we assume theta1 = 0, so we set tau1 = 0. theta2 is a nuisance parameter that we arbitrarily be N(0, 1). (The Bayes factor will be insensitive to this choice.)

5. Under A (not independence), we assume theta1 is N(0, tau1), where tau1 reflects our beliefs about the location of theta1 when the proportions are different. We also assume again that theta2 is N(0, 1). (This means that our beliefs about theta2 are insensitive to our beliefs about theta1.)

To compute the marginal densities, we write a function that computes the logarithm of the posterior when (theta1, theta2) have the above prior.

logctable.test=function (theta, datapar)
{
theta1 = theta[1] # log odds ratio
theta2 = theta[2] # log odds product

s1 = datapar$data[1,1]
f1 = datapar$data[1,2]
s2 = datapar$data[2,1]
f2 = datapar$data[2,2]

logitp1 = (theta1 + theta2)/2
logitp2 = (theta2 - theta1)/2
loglike = s1 * logitp1 - (s1 + f1) * log(1 + exp(logitp1))+
s2 * logitp2 - (s2 + f2) * log(1 + exp(logitp2))
logprior = dnorm(theta1,mean=0,sd=datapar$tau1,log=TRUE)+
dnorm(theta2,mean=0,sd=datapar$tau2,log=TRUE)

return(loglike+logprior)
}

We enter the data as a 2 x 2 matrix:

data=matrix(c(83,3,72,14),c(2,2))
data
[,1] [,2]
[1,] 83 72
[2,] 3 14

The argument datapar in the function is a list consisting of data, the 2 x 2 data table, and the values of tau1 and tau2.

Suppose we assume theta1 is N(0, .8) under the alternative hypothesis. This prior is shown in the below figure.

By using the laplace function, we compute the log marginal density under both models. (For H, we are approximating the point mass of theta1 on 1 by a normal density with a tiny standard deviation tau1.)

l.marg0=laplace(logctable.test,c(0,0),list(data=data,tau1=.0001,tau2=1))$int
l.marg1=laplace(logctable.test,c(0,0),list(data=data,tau1=0.8,tau2=1))$int

We compute the Bayes factor in support of the hypothesis A.

BF.10=exp(l.marg1-l.marg0)
BF.10
[1] 7.001088

The conclusion is that the alternative hypothesis A is seven times more plausible than the null hypothesis H.

Simple Illustration of Bayes Factors

Suppose we collect the number of accidents in a year for 30 Belgium drivers. We assume that y1,..., y30 are independent Poisson(lambda), where lambda is the average number of accidents for all Belgium drivers.

Consider the following four priors for lambda:

PRIOR 1: lambda is gamma with shape 3.8 and rate 8.1. This prior reflects the belief that the quartiles of lambda are 0.29 and 0.60.

PRIOR 2: lambda is gamma with shape 3.8 and rate 4. The mean of this prior is 0.95 so this prior reflects one's belief that lambda is close to 1.

PRIOR 3: lambda is gamma with shape 0.38 and 0.81. This prior has the same mean as PRIOR 1, but it is much more diffuse, reflecting weaker information about lambda.

PRIOR 4: log lambda is normal with mean -0.87 and standard deviation 0.60. On the surface, this looks different from the previous priors, but this prior also matches the belief that the quartiles of lambda are 0.29 and 0.60.

Suppose we observe some data -- for the 30 drivers, 22 had no accidents, 7 had exactly one accident, and 1 had two accidents. The likelihood is given by

LIKE = exp(-30 lambda) lambda^9

In the below graphs, we display the likelihood in blue and show the four priors in red. Here's the R code to produce one of the graphs. We simulate draws from the likelihood and the prior and display density estimates.

like=rgamma(10000,shape=10,rate=30)
p1=rgamma(10000,shape=3.8,rate=8.1)
plot(density(p1),xlim=c(0,3),ylim=c(0,4),
main="PRIOR 1",xlab="LAMBDA",lwd=3,col="red",col.main="red")
lines(density(like),lwd=3,col="blue")
text(1.2,3,"LIKELIHOOD",col="blue")

Note that Priors 1 and 4 are pretty consistent with the likelihood. There is some conflict between Prior 2 and the likelihood and Prior 3 is pretty flat relative to the likelihood.

We can compare the four models by use of Bayes factors. We first compute a function that computes the log posterior for each prior. There already is a function logpoissgamma in the LearnBayes package that computes the posterior of log lambda with Poisson sampling and a gamma prior. (This can be used for priors 1, 2, and 3.) The function logpoissnormal can be used for Poisson sampling and a normal prior (prior 4). Then we use the function laplace to approximate the value of the log predictive density.

For example, here's the code to compute the log marginal density for prior 1.

datapar=list(data=d,par=c(3.8,8.1))
laplace(logpoissgamma,.5,datapar)$int
0.4952788

So log m(y) for prior 1 is about 0.5.

We do this for each prior and get the following values:

model log m(y)
-----------------
PRIOR 1 0.495
PRIOR 2 -0.729
PRIOR 3 - 0.454
PRIOR 4 0.558

We can use this output to compute Bayes factors. For example, the Bayes factor in support of PRIOR 1 over PRIOR 2 is

BF_12 = exp(0.495 - (-0.729)) = 3.4

This means that the model with PRIOR 1 is about three and a half times as likely as the model with PRIOR 2. This is not surprising, seeing the conflict between the likelihood and the Bayes factor in the graph.

Conflict between Bayesian and Frequentist Measures of Evidence

Here's a simple illustration of the conflict between a p-value and a Bayesian measure of evidence.

Suppose you take a sample y1,..., yn from a normal population with mean mu and known standard deviation sigma. You wish to test

H: mu = mu0 A: mu not equal to mu0

The usual test is based on the statistic Z = sqrt(n)*(ybar - mu0)/sigma. One computes the p-value

p-value = 2 x P(Z >= z0)

and rejects H if the p-value is small. Suppose mu0 = 0, sigma = 1, and one takes a sample of size n = 4 and observe ybar = 0.98. Then one computes

Z = sqrt(4)*0.98 = 1.96

and the p-value is

p-value = 2 * P(Z > = 1.96) = 0.05.

Consider the following Bayes test of H against A. A Bayesian model is a specification of the sampling density and the prior density. One model M0 says that the mean mu = mu0. To complete the second model M1, we place a normal prior with mean mu0 and standard deviation tau on mu. The Bayes factor in support of the M0 over the model M1 is given by the ratio of predictive densities

BF = m(y | M0)/m(y|M1)

and the posterior probability of M0 is given by

P(M0| y) = p0 BF/(p0 BF + p1),

where p0 is the prior probability of M0.

The function mnormt.twosided in the LearnBayes package does this calculation. To use this function, we specify (1) the value m0 to be tested, (2) the prior probability of H, (3) the value of tau (the spread of the prior under A), and (4) the data vector that is (ybar, n, sigma).

Here we specify the inputs:

mu0=0; prob=.5; tau=0.5
ybar = 0.98; n = 4; sigma=1
data=c(ybar,n,sigma)

Then we can use mnormt.twosided -- the outputs are the Bayes factor and the posterior probability of H:

mnormt.twosided(mu0,prob,tau,data)
$bf
[1] 0.5412758

$post
[1] 0.3511868

We see that the posterior probability of H0 is 0.35 which is substantially higher than the p-value of 0.05.

In this calculation, we assumed that tau = 0.5 -- this reflect our belief about the spread of mu about mu0 under the alternative hypothesis. What if we chose a different value for tau?

We investigate the sensitivity of this posterior probability calculation with respect to tau.

We write a function that computes the posterior probability for a given value of tau.

post.prob=function(tau)
{
data=c(.98,4,1); mu0=0; prob0=.5
mnormt.twosided(mu0,prob,tau,data)$post
}

Then we use the curve function to plot this function for values of tau between 0.01 to 4.

curve(post.prob,from=.01,to=4,xlab="TAU",ylab="PROB(H0)",lwd=3,col="red")

In the figure below, it looks like the probability of H exceeds 0.32 for all tau.