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
102
103
104
105
106
107
108
109
110
111
112
|
% file MASS/man/housing.Rd
% copyright (C) 1999 W. N. Venables and B. D. Ripley
%
\name{housing}
\alias{housing}
\title{
Frequency Table from a Copenhagen Housing Conditions Survey
}
\description{
The \code{housing} data frame has 72 rows and 5 variables.
}
\usage{
housing
}
\format{
\describe{
\item{\code{Sat}}{
Satisfaction of householders with their present housing
circumstances, (High, Medium or Low, ordered factor).
}
\item{\code{Infl}}{
Perceived degree of influence householders have on the
management of the property (High, Medium, Low).
}
\item{\code{Type}}{
Type of rental accommodation, (Tower, Atrium, Apartment, Terrace).
}
\item{\code{Cont}}{
Contact residents are afforded with other residents, (Low, High).
}
\item{\code{Freq}}{
Frequencies: the numbers of residents in each class.
}
}
}
\source{
Madsen, M. (1976)
Statistical analysis of multiple contingency tables. Two examples.
\emph{Scand. J. Statist.} \bold{3}, 97--106.
Cox, D. R. and Snell, E. J. (1984)
\emph{Applied Statistics, Principles and Examples}.
Chapman & Hall.
}
\references{
Venables, W. N. and Ripley, B. D. (2002)
\emph{Modern Applied Statistics with S.} Fourth edition. Springer.
}
\examples{
options(contrasts = c("contr.treatment", "contr.poly"))
# Surrogate Poisson models
house.glm0 <- glm(Freq ~ Infl*Type*Cont + Sat, family = poisson,
data = housing)
summary(house.glm0, cor = FALSE)
addterm(house.glm0, ~. + Sat:(Infl+Type+Cont), test = "Chisq")
house.glm1 <- update(house.glm0, . ~ . + Sat*(Infl+Type+Cont))
summary(house.glm1, cor = FALSE)
1 - pchisq(deviance(house.glm1), house.glm1$df.residual)
dropterm(house.glm1, test = "Chisq")
addterm(house.glm1, ~. + Sat:(Infl+Type+Cont)^2, test = "Chisq")
hnames <- lapply(housing[, -5], levels) # omit Freq
newData <- expand.grid(hnames)
newData$Sat <- ordered(newData$Sat)
house.pm <- predict(house.glm1, newData,
type = "response") # poisson means
house.pm <- matrix(house.pm, ncol = 3, byrow = TRUE,
dimnames = list(NULL, hnames[[1]]))
house.pr <- house.pm/drop(house.pm \%*\% rep(1, 3))
cbind(expand.grid(hnames[-1]), round(house.pr, 2))
# Iterative proportional scaling
loglm(Freq ~ Infl*Type*Cont + Sat*(Infl+Type+Cont), data = housing)
# multinomial model
library(nnet)
(house.mult<- multinom(Sat ~ Infl + Type + Cont, weights = Freq,
data = housing))
house.mult2 <- multinom(Sat ~ Infl*Type*Cont, weights = Freq,
data = housing)
anova(house.mult, house.mult2)
house.pm <- predict(house.mult, expand.grid(hnames[-1]), type = "probs")
cbind(expand.grid(hnames[-1]), round(house.pm, 2))
# proportional odds model
house.cpr <- apply(house.pr, 1, cumsum)
logit <- function(x) log(x/(1-x))
house.ld <- logit(house.cpr[2, ]) - logit(house.cpr[1, ])
(ratio <- sort(drop(house.ld)))
mean(ratio)
(house.plr <- polr(Sat ~ Infl + Type + Cont,
data = housing, weights = Freq))
house.pr1 <- predict(house.plr, expand.grid(hnames[-1]), type = "probs")
cbind(expand.grid(hnames[-1]), round(house.pr1, 2))
Fr <- matrix(housing$Freq, ncol = 3, byrow = TRUE)
2*sum(Fr*log(house.pr/house.pr1))
house.plr2 <- stepAIC(house.plr, ~.^2)
house.plr2$anova
}
\keyword{datasets}
|
