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
113
114
115
116
117
118
119
120
121
122
123
124
|
\name{rsmodel}
\alias{rsmodel}
\alias{RSModel.fit}
\alias{print.rsmodel}
\alias{summary.rsmodel}
\alias{print.summary.rsmodel}
\alias{coef.rsmodel}
\alias{bread.rsmodel}
\alias{estfun.rsmodel}
\alias{logLik.rsmodel}
\alias{vcov.rsmodel}
\title{Rating Scale Model Fitting Function}
\description{
\code{rsmodel} is a basic fitting function for rating scale models.
}
\usage{
rsmodel(y, weights = NULL, start = NULL, reltol = 1e-10,
deriv = c("sum", "diff"), hessian = TRUE,
maxit = 100L, full = TRUE, \dots)
}
\arguments{
\item{y}{item response object that can be coerced (via \code{\link[base]{as.matrix}})
to a numeric matrix with scores 0, 1, \dots Typically, either
already a matrix, data frame, or dedicated object of class
\code{\link{itemresp}}.}
\item{weights}{an optional vector of weights (interpreted as case
weights).}
\item{deriv}{character. If "sum" (the default), the first derivatives
of the elementary symmetric functions are calculated with the sum
algorithm. Otherwise ("diff") the difference algorithm (faster but
numerically unstable) is used.}
\item{start}{an optional vector of starting values.}
\item{hessian}{logical. Should the Hessian of the final model be computed?
If set to \code{FALSE}, the \code{vcov} method can only return \code{NA}s
and consequently no standard errors or tests are available in the
\code{summary}.}
\item{reltol, maxit, \dots}{further arguments passed to \code{\link[stats]{optim}}.}
\item{full}{logical. Should a full model object be returned? If set to \code{FALSE},
no variance-covariance matrix and no matrix of estimating functions are computed.}
}
\details{
\code{rsmodel} provides a basic fitting function for rating scales models,
intended as a building block for fitting rating scale trees. It
estimates the rating scale model in the parametrization suggested by
Andrich (1978), i.e., item-specific parameters \eqn{\xi_{j}} who mark
the location of the first absolute threshold of an item on the theta axis and
cumulative relative threshold parameters \eqn{\kappa_{k}} are
estimated by the function \code{rsmodel}.
\code{rsmodel} returns an object of class \code{"rsmodel"} (and
class \code{"pcmodel"}) for which several basic methods are available,
including \code{print}, \code{plot}, \code{summary}, \code{coef},
\code{vcov}, \code{logLik}, \code{\link{discrpar}}, \code{estfun},
\code{\link{itempar}}, \code{\link{threshpar}}, and \code{\link{personpar}}.
}
\value{
\code{rsmodel} returns an S3 object of class \code{"rsmodel"},
i.e., a list with the following components:
\item{coefficients}{a named vector of estimated item-specific
parameters (without the first item parameter which is constrained
to 0) and estimated cumulative relative treshold parameters
(again without first threshold parameter which is also constrained to 0),}
\item{vcov}{covariance matrix of the parameters in the model,}
\item{data}{modified data, used for model-fitting, i.e., cleaned for items without
variance, centralized so that the first category is zero for all items
and without observations with zero weight. Be careful, this is different than for
objects of class \code{"raschmodel"} or \code{"btmodel"}, where
\code{data} contains the \emph{original} data,}
\item{items}{logical vector of length \code{ncol(y)}, which
indicates which items have variance (\code{TRUE}), i.e., are identified and have been
used for the estimation or not (\code{FALSE}),}
\item{categories}{integer vector of length \code{ncol(y)}, which
contains the number of categories minus one per item,}
\item{n}{number of observations (with non-zero weights),}
\item{n_org}{original number of observations in \code{y},}
\item{weights}{the weights used (if any),}
\item{na}{logical indicating whether the data contains NAs,}
\item{esf}{list of elementary symmetric functions and their
derivatives for estimated parameters,}
\item{loglik}{log-likelihood of the fitted model,}
\item{df}{number of estimated parameters,}
\item{code}{convergence code from \code{optim},}
\item{iterations}{number of iterations used by \code{optim},}
\item{reltol}{tolerance passed to \code{optim}.}
}
\references{
Andrich D (1978).
Application of a Psychometric Rating Model to Ordered Categories Which Are Scored with Successive Integers.
\emph{Psychometrika}, \bold{2}(4), 581--594.
}
\seealso{\code{\link{pcmodel}}, \code{\link{gpcmodel}}, \code{\link{raschmodel}},
\code{\link{plmodel}}, \code{\link{btmodel}}}
\examples{
o <- options(digits = 4)
## Verbal aggression data
data("VerbalAggression", package = "psychotools")
## Rating scale model for the other-to-blame situations
rsm <- rsmodel(VerbalAggression$resp[, 1:12])
summary(rsm)
## visualizations
plot(rsm, type = "profile")
plot(rsm, type = "regions")
plot(rsm, type = "curves")
plot(rsm, type = "information")
plot(rsm, type = "piplot")
options(digits = o$digits)
}
\keyword{regression}
|
