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
|
\name{relrelimp}
\alias{relrelimp}
\title{Comparison of Relative Importances in a
Multinomial Logit Model}
\description{
Produces a summary
of the relative importance of two predictors or two sets of predictors
in a fitted \code{\link[nnet]{multinom}} model object, and compares
relative importances
across two of the fitted logit models.
}
\usage{
relrelimp(object, set1=NULL, set2=NULL, label1="set1", label2="set2",
subset=TRUE,
response.cat1=NULL, response.cat2=NULL)
}
\arguments{
\item{object}{A model object of class
\code{multinom}}
\item{set1}{An index or vector of indices for the effects to be
included in the numerator of the comparison}
\item{set2}{An index or vector of indices for the effects to be
included in the denominator of the comparison}
\item{label1}{A character string; mnemonic name for the
variables in \code{set1}}
\item{label2}{A character string; mnemonic name for the
variables in \code{set2}}
\item{subset}{Either a vector of numeric indices for the cases to be included
in the standardization of effects, or a vector of logicals
(\code{TRUE} for inclusion)
whose length is the same as the number of rows in the model frame,
\code{object$model}.
The default choice is to include all cases in the model frame.}
\item{response.cat1}{A character
string used to specify the first regression of interest
(i.e., the regression
which predicts the log odds on \code{response.cat1} versus the model's
reference category). The \code{response.cat1} argument should be an
element of \code{object$lab}.}
\item{response.cat2}{A character
string used to specify the second regression of interest
(i.e., the regression
which predicts the log odds on \code{response.cat2} versus the model's
reference category). The \code{response.cat2} argument should be an
element of \code{object$lab}.}
}
\details{Computes a relative importance summary as described in
\code{\link{relimp}}, for each of the two regressions specified by
\code{response.cat1}
and \code{response.cat2} (relative to the same
reference category); and computes the
difference of those two relative importance summaries,
along with an estimated
standard error for that difference.
}
\value{
An object of class \code{relrelimp}, with at least the following components:
\item{model}{The call used to construct the model object summarized}
\item{sets}{The two sets of indices specified as arguments}
\item{response.category}{A character vector containing the specified
\code{response.cat1} and \code{response.cat2}}
\item{log.ratio}{The natural logarithm of the ratio of effect
standard deviations corresponding to the two sets specified.
A vector with
three components: the first is for \code{response.cat1}
versus the reference
category, the second for \code{response.cat2} versus the
reference category,
the third is the difference.}
\item{se.log.ratio}{Estimated standard errors for the elements of
\code{log.ratio}}
}
\author{David Firth, \email{d.firth@warwick.ac.uk} }
\seealso{\code{\link{relimp}}}
\examples{
## Data on housing and satisfaction, from Venables and Ripley
library(MASS)
library(nnet)
data(housing)
house.mult <- multinom(Sat ~ Infl + Type + Cont, weights = Freq,
data = housing)
relrelimp(house.mult, set1 = 2:3, set2 = 7,
label1 = "Influence", label2 = "Contact",
response.cat1 = "Medium", response.cat2 = "High")
## Computes the relative contribution of Influence and Contact in
## each of the two logistic regressions (Med/Low and High/Low), and
## compares those two relative-contribution measures.
}
\keyword{models}
\keyword{regression}
|
