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
|
\name{btmodel}
\alias{btmodel}
\alias{btReg.fit}
\alias{print.btmodel}
\alias{summary.btmodel}
\alias{print.summary.btmodel}
\alias{coef.btmodel}
\alias{worth.btmodel}
\alias{deviance.btmodel}
\alias{logLik.btmodel}
\alias{vcov.btmodel}
\alias{estfun.btmodel}
\title{Bradley-Terry Model Fitting Function}
\description{
\code{btmodel} is a basic fitting function for simple Bradley-Terry models.
}
\usage{
btmodel(y, weights = NULL, type = c("loglin", "logit"), ref = NULL,
undecided = NULL, position = NULL, start = NULL, vcov = TRUE, estfun =
FALSE, \dots)
}
\arguments{
\item{y}{paircomp object with the response.}
\item{weights}{an optional vector of weights (interpreted as case weights).}
\item{type}{character. Should an auxiliary log-linear Poisson model or logistic
binomial be employed for estimation? The latter is not available if
undecided effects are estimated.}
\item{ref}{character or numeric. Which object parameter should be the reference
category, i.e., constrained to zero?}
\item{undecided}{logical. Should an undecided parameter be estimated?}
\item{position}{logical. Should a position effect be estimated?}
\item{start}{numeric. Starting values when calling \code{\link[stats]{glm.fit}}.}
\item{vcov}{logical. Should the estimated variance-covariance be included in
the fitted model object?}
\item{estfun}{logical. Should the empirical estimating functions (score/gradient
contributions) be included in the fitted model object?}
\item{\dots}{further arguments passed to functions.}
}
\details{
\code{btmodel} provides a basic fitting function for Bradley-Terry models,
intended as a building block for fitting Bradley-Terry trees and
Bradley-Terry mixtures in the \pkg{psychotree} package, respectively. While
\code{btmodel} is intended for individual paired-comparison data, the
\pkg{eba} package provides functions for aggregate data.
\code{btmodel} returns an object of class \code{"btmodel"} for which
several basic methods are available, including \code{print}, \code{plot},
\code{summary}, \code{coef}, \code{vcov}, \code{logLik}, \code{estfun}
and \code{\link{worth}}.
}
\value{
\code{btmodel} returns an S3 object of class \code{"btmodel"},
i.e., a list with components as follows.
\item{y}{paircomp object with the response}
\item{coefficients}{estimated parameters on log-scale (without the first
parameter which is always constrained to be 0),}
\item{vcov}{covariance matrix of the parameters in the model,}
\item{loglik}{log-likelihood of the fitted model,}
\item{df}{number of estimated parameters,}
\item{weights}{the weights used (if any),}
\item{n}{number of observations (with non-zero weights),}
\item{type}{character for model type (see above),}
\item{ref}{character for reference category (see above),}
\item{undecided}{logical for estimation of undecided parameter (see above),}
\item{position}{logical for estimation of position effect (see above),}
\item{labels}{character labels of the objects compared,}
\item{estfun}{empirical estimating function (also known as scores or gradient contributions).}
}
\seealso{\code{\link{pcmodel}}, \code{\link{gpcmodel}}, \code{\link{rsmodel}},
\code{\link{raschmodel}}, \code{\link{plmodel}}, the \pkg{eba} package}
\examples{
o <- options(digits = 4)
## data
data("GermanParties2009", package = "psychotools")
## Bradley-Terry model
bt <- btmodel(GermanParties2009$preference)
summary(bt)
plot(bt)
options(digits = o$digits)
}
\keyword{regression}
|
