1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
|
\name{inter_stats}
\alias{inter_stats}
\title{
Compute Posterior Summary Statistics of the Log-Linear Parameters.
}
\description{
This function computes the posterior summary statistics of the log-linear parameters
using the MCMC output of \code{"bcct"} and \code{"bict"} objects. The posterior summary
statistics are posterior probability, posterior mean, posterior variance and lower and
upper limits highest posterior density intervals (HPDIs).
}
\usage{
inter_stats(object, cutoff = 0.75, n.burnin = 0, thin = 1, prob.level = 0.95)
}
\arguments{
\item{object}{
An object of class \code{"bcct"} or \code{"bict"}.
}
\item{cutoff}{
An optional argument giving the cutoff posterior probability for displaying posterior
summary statistics of the log-linear parameters. Only those log-linear parameters with
a posterior probability greater than \code{cutoff} will be returned as part of the output.
The default value is 0.75.
}
\item{n.burnin}{
An optional argument giving the number of iterations to use as burn-in.
The default value is 0.
}
\item{thin}{
An optional argument giving the amount of thinning to use, i.e. the computations are
based on every \code{thin}-th value in the MCMC sample. The default value is 1, i.e. no
thinning.
}
\item{prob.level}{
An optional argument giving the probability content of the HPDIs. The default value is 0.95.
}
}
\details{
This function provides an expanded version of what \code{\link{inter_probs}} provides.
The use of thinning is recommended when the number of MCMC iterations and/or the number of
log-linear parameters in the maximal model are/is large, which may cause problems with
comuter memory storage.
}
\value{
This function will return an object of class \code{"interstat"} which is a list with the following
components:
\item{term}{A vector of term labels for each parameter.}
\item{prob}{A vector of posterior probabilities for each parameter.}
\item{post_mean}{A vector of posterior means for each parameter.}
\item{post_var}{A vector of posterior variances for each parameter.}
\item{lower}{A vector of lower limits for the 100*\code{prob.level}\% HPDI for each parameter.}
\item{upper}{A vector of upper limits for the 100*\code{prob.level}\% HPDI for each parameter.}
\item{prob.level}{The argument \code{prob.level}.}
The function will only return elements in the above list if \code{prob} > \code{cutoff}.
}
\author{
Antony M. Overstall \email{A.M.Overstall@soton.ac.uk}.
}
\seealso{
\code{\link{bcct}},
\code{\link{bict}},
\code{\link{print.interstat}}
\code{\link{inter_probs}}
}
\examples{
set.seed(1)
## Set seed for reproducibility
data(AOH)
## Load AOH data
test1<-bcct(formula=y~(alc+hyp+obe)^3,data=AOH,n.sample=100,prior="UIP")
## Starting from maximal model of saturated model do 100 iterations of MCMC
## algorithm.
inter_stats(test1,n.burnin=10,cutoff=0.5)
## Calculate posterior summary statistics having used a burn-in phase of
## 10 iterations and a cutoff of 0 (i.e. display all terms with
## non-zero posterior probability. Will get the following:
#Posterior summary statistics of log-linear parameters:
# post_prob post_mean post_var lower_lim upper_lim
#(Intercept) 1 2.88291 0.002565 2.78778 2.97185
#alc1 1 -0.05246 0.008762 -0.27772 0.06655
#alc2 1 -0.05644 0.006407 -0.20596 0.11786
#alc3 1 0.06822 0.005950 -0.09635 0.18596
#hyp1 1 -0.53895 0.003452 -0.63301 -0.39888
#obe1 1 -0.04686 0.007661 -0.20929 0.12031
#obe2 1 0.01395 0.004024 -0.11024 0.11783
#NB: lower_lim and upper_lim refer to the lower and upper values of the
#95 % highest posterior density intervals, respectively
}
|
