\documentclass[nojss]{jss}
\usepackage{amsmath,amsbsy,thumbpdf,lmodern}
%\VignetteIndexEntry{equateIRT: An R Package for IRT Test Equating}
\def\ba{\boldsymbol\alpha}
\def\gg{{g-1, \, g}}
\def\g{{g-1}}
\def\acov{\mathsf{ACOV}}
\def\pth{\mathcal{P}}
\def\bA{{\bf A}}
\def\bB{{\bf B}}
\author{Michela Battauz\\Universit\`a degli Studi di Udine}
\Plainauthor{Michela Battauz}
\title{\pkg{equateIRT}: An \proglang{R} Package for IRT Test Equating}
\Plaintitle{equateIRT: An R Package for IRT Test Equating}
\Shorttitle{\pkg{equateIRT}:IRT Test Equating in \proglang{R}}
\Abstract{
This introduction to the \proglang{R} package \pkg{equateIRT} is a
(slightly) modified version of \citet{batt:equa:2015}, published in the
\emph{Journal of Statistical Software}.
The \proglang{R} package \pkg{equateIRT} implements item response theory (IRT)
methods for equating different forms
composed of dichotomous items. In particular, the IRT models included are
the three-parameter logistic model, the two-parameter logistic model,
the one-parameter logistic model and the Rasch model.
Forms can be equated when they present common items (direct equating)
or when they can be linked through a chain of forms that present common items in pairs (indirect or chain equating).
When two forms can be equated through different paths, a single conversion can be obtained by averaging
the equating coefficients.
The package calculates direct and chain equating coefficients. The averaging of direct and chain
coefficients that link the same two forms is performed through the bisector method.
Furthermore, the package provides analytic standard errors of direct, chain and average equating coefficients.
}
\Keywords{bisector, chain, equating, equating coefficients, IRT, Rasch model, standard errors}
\Address{
Michela Battauz\\
Department of Economics and Statistics\\
Universit\`a degli Studi di Udine\\
via Tomadini 30/A\\
33100 Udine, Italy\\
E-mail: \email{michela.battauz@uniud.it}\\
URL: \url{https://people.uniud.it/page/michela.battauz}
}
\begin{document}
\section{Introduction}
In many testing programs, security reasons require that test forms are
composed of different items, making test scores not comparable across
different administrations. The equating process permits the
comparison of scores obtained from different forms taken. The term
equating traditionally refers to the adjustment of scores from
parallel forms, that are as similar as possible in content and
statistical characteristics \citep[][Chapter~1.1.2 and Chapter~1.2.3]{Kole:Bren:test:2014}. The methods presented in this paper,
implemented in the \proglang{R} \citep{erre} package \pkg{equateIRT}
\citep{equateIRT}, allow more general forms of equating, such as
horizontal and vertical scaling. For example, vertical scaling is
intended to make scores comparable across different educational
grades, where the content of the tests differs accordingly to the
educational level. The package is available from the Comprehensive
\proglang{R} Archive Network (CRAN) at
\url{http://CRAN.R-project.org/package=equateIRT}.
Various statistical methods have been proposed to perform equating
\citep[for a broad review, see][]{Kole:Bren:test:2014}. The
\proglang{R} package \pkg{equateIRT} focuses on item response theory
(IRT) methods for dichotomous items. IRT models are statistical
models that have as response variable the responses given to the items
of a questionnaire or test. These responses are modeled as a function
of a latent trait, for example the ability level, and the item
parameters that are related to some characteristics of the items, such
as the difficulty. The purpose of IRT models is to provide a measure
of the latent trait under investigation. For a broad review of IRT
models see \citet{Lind:Hamb:hand:1997}. The IRT models included in
the \pkg{equateIRT} package are the three-parameter logistic model,
the two-parameter logistic model, the one-parameter logistic model and
the Rasch model. When the parameters of the IRT model are estimated
separately for each group of people taking a different test form, they
are not comparable because the origin of the measurement scale is not
identifiable. However, when two forms present a subset of common
items, the parameters can be put on the same scale. There are two
methods available to pursue this end. Concurrent calibration consists
of estimating the item parameters together and yields measurements
directly on a common metric. Alternatively, item parameters are
estimated separately for each test form (separate calibration) and the
estimates of the item parameters for the common items can be used to
estimate the scale transformation. Both approaches can be extended to
the case of multiple test forms, provided that the forms can be linked
through common items. Concurrent calibration presents the advantage
of not requiring any conversion after the estimation of the parameters
as they are already expressed on the same scale. However, this
approach requires the combination of the data from each form in a
single dataset that could become quite large when there are many
forms or when each form is administered to a very large number of
people. In some cases, the large dimension of the dataset then makes
the estimation very slow or even unfeasible. Instead, by equating the
coefficients of separate calibrations, it is not necessary to estimate
again the parameters of previous administrations, thus avoiding the
construction of a single dataset.
The conversion of parameter estimates is attained by applying a linear transformation and
the coefficients of this transformation are called equating coefficients.
For each pair of forms containing common items, direct equating coefficients can be calculated.
Suppose that Forms~1 and 2 have common items and that Forms~2 and 3 have common items.
Forms~1 and 3 can then be equated employing the chain going through Form 2. In this case,
indirect equating coefficients linking Forms~1 and 3 can be calculated as a function of direct
equating coefficients linking pairs of forms with common items. In general, when two forms can be
linked through a chain of forms that present common items in pairs, indirect equating coefficients
can be calculated that permit the conversion of parameter estimates of one form into the scale of
the other form using a linear transformation. These coefficients are a function of direct equating
coefficients. Furthermore, some linkage designs are quite complex and two forms can be linked
through different chains and possibly a direct link. For every path that links the same two forms
the equating coefficients can be computed, thus yielding different scale conversions. In this case,
the equating coefficients can be averaged in order to obtain a single transformation.
For the computation of direct equating coefficients, the \pkg{equateIRT} package implements methods
based on moments of item parameters, that are the mean-sigma \citep{Marc:item:1977},
the mean-mean \citep{Loyd:vert:1980}, and the mean-geometric mean \citep{Misl:bilo:1990}
methods, and response function methods, that are the Haebara \citep{Haeb:equa:1980}
and the Stocking-Lord \citep{Stoc:deve:1983} methods.
The package computes also indirect equating coefficients through a chain of forms.
The bisector method is used to average equating coefficients derived from different paths
\citep{Holl:howt:2011, batt:irtt:2013}.
Since the estimated equating coefficients are subject to sampling
variation, it is important to assess the magnitude of this variability
in order to evaluate the accuracy of the equating process
\citep{Ogas:appl:2011}. This objective can be accomplished by
observing the asymptotic standard errors of the equating coefficients.
Despite the importance of verifying the precision of the equating
performed, equating software generally does not provide standard
errors of equating coefficients. To the best of my knowledge, just a
few computer programs calculate standard errors based on bootstrap
techniques. The computation of analytic standard errors of IRT
equating coefficients is not implemented in any software, with the
only exception being a set of subroutines that implement the methods
developed by \citet{Ogas:asym:2000} and \citet{Ogas:stan:2001} for
direct equating coefficients, available from the author. Instead, the
\pkg{equateIRT} package provides analytical standard errors for
direct, chain and average equating coefficients. It is important to
note that analytical standard errors of equating coefficients can only
be computed if the covariance matrix of item parameter estimates is
available. This covariance matrix should be obtained from the software
used to estimate the item parameters. The package provides functions
to import data from \pkg{flexMIRT} \citep{flexmirt}, \pkg{IRTPRO}
\citep{irtpro} and the \proglang{R} packages \pkg{ltm} \citep{ltm} and
\pkg{mirt} \citep{mirt}. All of these IRT programs supply the
covariance matrix of item parameter estimates.
Other \proglang{R} packages provide implementation of IRT equating
methods. The package \pkg{irtoys} \citep{irtoys} performs IRT
equating for dichotomous items and the package \pkg{plink}
\citep{plink} implements unidimensional and multidimensional IRT
equating for both dichotomous and polytomous items. The \pkg{plink}
package performs chain equating by providing direct equating
coefficients between forms that present common items, but does not
provide indirect equating coefficients that permit the conversion from
the first form into the last form of the path. Chain equating is not
included in the \pkg{irtoys} package. Furthermore, the packages
\pkg{irtoys} and \pkg{plink} do not provide average equating
coefficients nor standard errors of direct, indirect and average
coefficients.
The paper is structured as follows. The theory on IRT test equating
is summarized in Section~\ref{IRTte}. Section~\ref{equateIRTpkg}
illustrates the \pkg{equateIRT} package and Section~\ref{con}
concludes the paper.
\section{IRT test equating}
\label{IRTte}
Consider a single test form that is denoted by $g$. In the
three-parameter logistic model, the probability of a positive response
on item $j$ in form $g$ for a person with ability $\theta$ is given by
\begin{equation}
\label{pr}
p_{gj}(\theta_{(g)}; a_{gj}, b_{gj}, c_{gj})=c_{gj}+(1-c_{gj})
\frac{\exp \left \{ D a_{gj}(\theta_{(g)}-b_{gj}) \right \} }
{1+\exp \left \{ D a_{gj}(\theta_{(g)}-b_{gj}) \right \} },
\end{equation}
where $a_{gj}$ is the item discrimination parameter, $b_{gj}$ is the item
difficulty parameter, $c_{gj}$ is the item guessing parameter and $D$ is a
constant typically set to 1.7.
We define the parameter vector of form $g$
as $\ba_g=(\ba_{g1}^\top,\dots ,\ba_{gn_g}^\top)^\top$,
where $\ba_{gj}=(a_{gj},b_{gj},c_{gj})^\top, j=1,\dots,n_g$,
and $n_g$ is the number of items of form $g$.
Item parameters are estimated separately for each form
by using the marginal maximum likelihood method
\citep{Bock:Aitk:marg:1981}, regarding the person parameter $\theta$ as a
random variable with standard normal distribution.
\clearpage
\subsection{Direct equating}
\label{deq}
Let $g-1$ be another form that presents $n_\gg$ items in common with Form $g$.
The parameters estimated for Form $g-1$ can be transformed to the scale
of Form $g$ by using the following equations
\begin{align}
\label{tet}
\theta_g &= A_\gg \, \theta_\g+B_\gg \, ,\\
\label{tea}
a_g &= \frac{a_\g}{A_\gg},\\
\label{teb}
b_g &= A_\gg \, b_\g+B_\gg \, ,
\end{align}
where $A_\gg$ and $B_\gg$ are the equating coefficients.
These coefficients can be estimated by using moments of
item parameters \citep[][Chapter~6.3.2]{Kole:Bren:test:2014},
or response function methods \citep[][Chapter~6.3.3]{Kole:Bren:test:2014}.
In any case, direct equating coefficients are estimated using only the item parameter estimates for
the items in common between two forms, irrespective of the items in common with other forms.
The mean-sigma, mean-mean and
mean-geometric mean methods are based on moments of item parameters.
In particular, the equating coefficient $A_\gg$ is given by
\begin{equation}
\label{ms}
A_\gg = \sqrt{\frac{\sum_{j=1}^{n_\gg} b_{gj}^2-\frac{1}{n_\gg}\left ( \sum_{j=1}^{n_\gg} b_{gj} \right )^2}{\sum_{j=1}^{n_\gg} b_{\g j}^2-\frac{1}{n_\gg}\left ( \sum_{j=1}^{n_\gg} b_{\g j} \right )^2}}
\end{equation}
using the mean-sigma method,
\begin{equation}
\label{mm}
A_\gg = \frac{\sum_{j=1}^{n_\gg} a_{\g j}}{\sum_{j=1}^{n_\gg} a_{gj}}
\end{equation}
using the mean-mean method, and
\begin{equation}
\label{mg}
A_\gg = \left ( \prod_{j=1}^{n_\gg} \frac{a_{\g j}}{a_{gj}} \right )^{\frac{1}{n_\gg}}
\end{equation}
using the mean-geometric mean method,
while the equating coefficient $B_\gg$ is given by
\begin{equation}
\label{b}
B_\gg = \frac{1}{n_\gg}\sum_{j=1}^{n_\gg} b_{gj}- A_\gg \; \frac{1}{n_\gg}\sum_{j=1}^{n_\gg} b_{\g j}
\end{equation}
for all methods.
The Haebara and Stocking-Lord methods are based
on the response function.
The equating coefficients using the Haebara method are obtained by minimizing the following function
\begin{equation}
\label{h}
f_H(A_\gg, B_\gg) = \frac{1}{2}\int_{-\infty}^{+\infty}\sum_{j=1}^{n_\gg} \left [ p_{gj}(\theta; a_{gj}, b_{gj}, c_{gj}) -
p_{\g j}(\theta; a^*_{\g j}, b^*_{\g j}, c_{\g j}) \right ]^2 h(\theta) d\theta,
\end{equation}
where $h(\theta)$ is the density of a standard normal variable, $a^*_{\g j}=a_{\g j}/A_\gg$ and
$b^*_{\g j}=A_\gg \, b_{\g j} + B_\gg$.
The Stocking-Lord method requires, instead, the minimization of the function
\begin{equation}
\label{sl}
f_{SL}(A_\gg, B_\gg) = \frac{1}{2}\int_{-\infty}^{+\infty}\left\{\sum_{j=1}^{n_\gg} \left [ p_{gj}(\theta; a_{gj}, b_{gj}, c_{gj}) -
p_{\g j}(\theta; a^*_{\g j}, b^*_{\g j}, c_{\g j}) \right ]\right\}^2 h(\theta) d\theta.
\end{equation}
The integrals in Equations~\ref{h} and~\ref{sl} do not have an analytical solution and they are generally
approximated using Gaussian quadrature.
Using the delta method, \citet{Ogas:asym:2000} and \citet{Ogas:stan:2001} derived
the asymptotic covariance matrix for
the vector of estimates $(\hat A_\gg,\hat B_\gg)^\top$, that is given by
\begin{equation}
\label{asc}
\acov(\hat A_\gg,\hat B_\gg)^\top=\frac{\partial (A_\gg,B_\gg)^\top}{\partial \ba_\gg^\top}
\acov(\hat \ba_\gg)\frac{\partial (A_\gg,B_\gg)}{\partial \ba_\gg} \, ,
\end{equation}
where $\ba_\gg=(\ba_g^\top,\ba_{g-1}^\top)^\top$ is a vector containing all the item parameters
related to Forms~$g-1$ and $g$,
$\partial (A_\gg,B_\gg)^\top/\partial \ba_\gg^\top$
is the matrix containing the partial derivatives of $A_\gg$ and $B_\gg$ with
respect to the item parameters evaluated at their true values,
and $\acov(\hat \ba_\gg)$ is the asymptotic covariance matrix
of $\hat \ba_\gg$.
The derivatives depend on the method used to determine the equating coefficients
and are given in \citet{Ogas:asym:2000, Ogas:appl:2011} for methods
based on moments, and in \citet{Ogas:stan:2001} for response function methods.
The estimate of the asymptotic covariance matrix is obtained
by inserting item parameter estimates into Equation~\ref{asc}.
\subsection{Equating chains}
\label{ech}
Suppose that two forms are linked through
a chain of forms that present
common items in pairs.
Define the path from Form $0$ to Form $l$
as $p=\{ 0,1,\dots,l \}$.
According to \citet{batt:irtt:2013},
it is possible to obtain
the equating coefficients transforming the
scale of $\theta_0$ to that of $\theta_l$ as a function of the direct equating coefficients
that link the forms with common items.
These coefficients will be referred to as indirect or chain equating
coefficients and they are given by
\begin{align}
A_p&=A_{0,1,\dots,l}=\prod_{g=1}^l A_{\gg}
\intertext{and}
B_p&=B_{0,1,\dots,l}=\sum_{g=1}^l B_{\gg} \; A_{g, \dots, l} \, ,
\end{align}
where $A_{g, \dots, l}=\prod_{h=g+1}^l A_{h-1, \, h}$ is the coefficient
that links Form $g$ to Form $l$.
Similarly to the case of a direct link,
the delta method can be exploited
to obtain the asymptotic covariance matrix
of the vector of estimates $(\hat A_p, \hat B_p)^\top$,
that is
\begin{equation}
\label{ind}
\acov(\hat A_p,\hat B_p)^\top=\frac{\partial (A_p,B_p)^\top}{\partial \ba_p^\top}
\acov(\hat \ba_p)\frac{\partial (A_p,B_p)}{\partial \ba_p},
\end{equation}
where $\ba_p=(\ba_0^\top,\ba_1^\top,\dots,\ba_l^\top)^\top$
is the vector containing all the item parameters
related to the forms that compose the path,
$\partial (A_p,B_p)^\top/\partial \ba_p^\top$
is the matrix containing the partial derivatives of $A_p$ and $B_p$ with
respect to the item parameters evaluated at their true values,
and $\acov(\hat \ba_p)$ is the asymptotic covariance
matrix of the estimate $\hat \ba_p$.
The derivatives are given in \citet{batt:irtt:2013}.
The estimate of the asymptotic covariance matrix is obtained
by inserting item parameter estimates into Equation~\ref{ind}.
The computation of chain equating coefficients is then performed in
two steps: first, the calculation of direct equating coefficients
between consecutive forms of a path and second, the determination of
indirect equating coefficients as a function of direct equating
coefficients. Thus, in this process, only the items in common between
two consecutive forms of a given chain are used. If, for example,
Forms~$0$ and $2$ present common items, this information is not used
to estimate the chain equating coefficients for path
$p=\{ 0,1,\dots,l \}$. This link can be used to constitute a further
path that connects Forms~$0$ and $l$ and a different scale conversion
can be derived for this path. In this case, there are two different
equatings for the same two forms, and they can be averaged as
explained in Section~\ref{aec}.
\subsection{Average equating coefficients}
\label{aec}
Suppose that two forms are linked through different paths.
Define the set of paths that link two
Forms~$0$ and $l$ as $\pth_{0l}$ and the linking coefficients
related to path $p$ as $A_p$ and $B_p$, $p\in\pth_{0l}$.
The equations that transform the scale of $\theta_0$ to that of $\theta_l$ are then
\begin{equation}
\label{lt}
\theta_l^{p}=A_{p}\;\theta_0+B_{p}, \quad p \in \pth_{0l}.
\end{equation}
As observed by \citet[][p.~280]{Kole:Bren:test:2014} and
\citet[p.~44]{Brau:obse:1982}, the equating relationships
provided by each path could be averaged to produce a single
conversion that is expected to be more accurate.
\citet{batt:irtt:2013} proposed using the bisector method,
suggested by \citet{Holl:howt:2011}
to average the equating functions obtained by using
different equating methods.
The angle bisector, in case of two linear functions that intersect
at a point, is the
linear function that bisects the angle between them.
The bisector method satisfies the symmetry property, which
requires that the inverse function of the average equating function equals
the average of the inverse functions.
Instead, the (weighted) mean of the equating functions does not satisfy the symmetry property,
making the bisector method preferable.
The bisector method yields a weighted average of the linear transformations~(\ref{lt}):
\begin{equation}
\label{ave}
\theta_l^{*} = \sum_{p\in\pth_{0l}} w_p \,\theta_l^p,
\end{equation}
where
\begin{equation}
w_p=\frac{n_p(1+A_p^2)^{-1/2}}{\sum_{b\in\pth_{0l}} n_b(1+A_b^2)^{-1/2}},
\end{equation}
and $n_p$ are optional weights associated with each path.
The average equating coefficients are then
\begin{equation}
A^{*}_{0l}=\sum_{p\in\pth_{0l}} A_p w_p
\end{equation}
and
\begin{equation}
B^{*}_{0l}=\sum_{p\in\pth_{0l}} B_p w_p.
\end{equation}
The asymptotic covariance matrix of the vector of estimates
$(\hat A^*_{0l}, \hat B^*_{0l})^\top$
can then be obtained by using the delta method as follows
\begin{equation}
\label{asc0l}
\acov(\hat A^*_{0l},\hat B^*_{0l})^\top=\frac{\partial (A^*_{0l},B^*_{0l})^\top}{\partial \ba^\top}
\acov(\hat \ba)\frac{\partial (A^*_{0l},B^*_{0l})}{\partial \ba},
\end{equation}
where $\ba=(\ba_p)_{p\in\pth_{0l}}$ is the vector containing all the item parameters
used in the equating process in at least one of the paths in $\pth_{0l}$,
$\partial (A^*_{0l},B^*_{0l})^\top/\partial \ba^\top$
is the matrix containing the partial derivatives of $A^*_{0l}$ and $B^*_{0l}$ with
respect to the item parameters evaluated at their true values,
and $\acov(\hat \ba)$ is a block diagonal matrix composed of all
$\acov(\hat \ba_p)$, $p\in\pth_{0l}$, and it is
the asymptotic covariance matrix of the
estimate $\hat \ba$.
The derivatives are given in \citet{batt:irtt:2013}.
The estimate of the asymptotic covariance matrix is obtained
by inserting item parameter estimates into Equation~\ref{asc0l}.
Obtaining average equating coefficients is then a process that takes place over three phases: First,
the estimation of direct equating coefficients, then the calculation of chain equating coefficients
for more than one path connecting two forms, and finally the computation of average equating
coefficients. The distribution of common items across forms determines which forms can be directly
linked in the first phase. From this follows the composition of paths connecting each pair of forms
and the computation of chain equating coefficients. There are no restrictions on these connections:
Different paths can share some parts, or the same item can be used to link more than one form.
The relationship between the average coefficients and the item parameters used to compute them
is reflected in the derivatives
$\frac{\partial (A^*_{0l},B^*_{0l})^\top}{\partial \ba^\top}$,
that determine the covariance matrix of the average equating coefficients.
A further issue concerns which weights $n_p$ to use in averaging.
\citet{batt:irtt:2013} proposed determining weights by
minimizing the average variance of $\theta^*_l$,
namely
\begin{equation}
\label{of}
\E_{\theta_0} \left [\VAR(\hat A^*_{0l} \, \theta_0 + \hat B^*_{0l} \big\vert\theta_0 ) \right ] =
\VAR(\hat A^*_{0l})+\VAR(\hat B^*_{0l}),
\end{equation}
assuming that $\theta_0$ has zero mean and variance equal to one.
\section[The equateIRT package]{The \pkg{equateIRT} package}
\label{equateIRTpkg}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
<>=
options(prompt = "R> ", continue = "+ ", width = 70, useFancyQuotes = FALSE)
@
In order to perform equating with the \pkg{equateIRT} package,
it is necessary to have previously estimated item parameters.
Item parameters can be estimated using \proglang{R} packages,
or using external software and
then import them into \proglang{R}.
To calculate standard errors of equating coefficients it is
also necessary to have estimated the covariance matrix of the item
parameter estimates. If the program used to estimate the IRT
model does not provide the covariance matrix of
item parameter estimates or the user is not interested in
standard errors of equating coefficients, the covariance
matrix can be set to \code{NULL}.
\subsection{Data preparation}
To the best of my knowledge, the computer programs for IRT model estimation
that export the covariance matrix of parameter estimates are \pkg{flexMIRT}, \pkg{IRTPRO} and
the \proglang{R} packages \pkg{ltm} and \pkg{mirt}.
The \pkg{equateIRT} package provides functions to import data from these programs.
Item parameters and the covariance matrices can be also imported from other programs.
In this case, the user should import item parameter estimates and the covariance matrices
using the general \proglang{R} function \code{read.table}.
For item parameter estimation, the programs \pkg{flexMIRT},
\pkg{IRTPRO} and the \proglang{R} packages \pkg{ltm} and \pkg{mirt}
formulate the three-parameter logistic model as a latent trait model
\citep{Barth:late:2011},
\begin{equation}
\label{ltpar}
\pi_j=c_j + (1 - c_j) \frac{\exp(\beta_{1j} + \beta_{2j} z)}{1 + \exp(\beta_{1j} + \beta_{2j} z)},
\end{equation}
where $\pi_j$ is the probability of a positive response for the $j$th item,
$\beta_{j1} = - D \cdot a_j \cdot b_j$ are easiness parameters, $\beta_{2j} = D \cdot a_j$ are discrimination
parameters, $c_j$ are guessing parameters and $z = \theta$ the latent ability.
In the following, this parameterization will be referred to as the \emph{latent trait parameterization}.
The two-parameter logistic model, the one-parameter logistic model and the Rasch model can be presented with the same formulation.
The two-parameter logistic model can be obtained by setting $c_j = 0$, the one-parameter logistic model
can be obtained by constraining also $\beta_{2j}$ to be constant across items, and the Rasch model can be obtained by setting $c_j = 0$ and $\beta_{2j}=1$.
Furthermore, for a three-parameter logistic model, if the guessing parameters are given under the parameterization
\begin{equation}
\label{lpar}
c_j = \frac{\exp(c^*_j)}{1+\exp(c^*_j)},
\end{equation}
this will be called the \emph{logistic parameterization} in the following.
The IRT programs estimate the parameters $(c_j^*, \beta_{1j}, \beta_{2j})$,
for every item $j$, and then calculate the item parameters of the usual IRT parameterization
given in Equation~\ref{pr} using Equation~\ref{lpar} and the equations
\begin{equation}
b_j=-\frac{\beta_{1j}}{D a_j} \qquad \textrm{and} \qquad a_j=\frac{\beta_{2j}}{D}.
\end{equation}
The covariance matrix of parameter estimates exported by the programs is related to
the vector of parameters $(c_j^*, \beta_{1j}, \beta_{2j})$. The covariance matrix of
$(c_j, b_j, a_j)$ can be obtained on the basis of the covariance matrix provided by the IRT programs
by applying the delta method. The \pkg{equateIRT} package provides this functionality.
The functions provided by the \pkg{equateIRT} package to import item parameter estimates and
the covariance matrix are \code{import.ltm}, \code{import.mirt}, \code{import.flexmirt} and \code{import.irtpro}.
Since \pkg{ltm} and \pkg{mirt} are \proglang{R} packages, functions \code{import.ltm} and \code{import.mirt}
just extract item parameter
estimates and the covariance matrix from an object previously created with these packages.
The arguments of the functions are:
\begin{description}
\item[\code{mod}:] Output object from functions \code{rasch}, \code{ltm}, or \code{tpm}
of the \pkg{ltm} package, or from function \code{mirt} of the \pkg{mirt} package.
\item[\code{display}:] Logical value indicating whether the coefficients and the standard errors
should be printed. The default is \code{TRUE}.
\item[\code{digits}:] Integer value indicating the number of decimal digits to be used if display is \code{TRUE}.
\end{description}
Instead, functions \code{import.flexmirt} and \code{import.irtpro} read external files
previously created with the programs \pkg{flexMIRT} or \pkg{IRTPRO}.
The arguments of the functions are:
\begin{description}
\item[\code{fnamep}:] The name of the file containing the item parameter estimates.
This is a file whose name ends with ``-prm.txt''.
\item[\code{fnamev}:] The name of the file containing the covariance matrix of item parameter estimates.
This is a file ending with ``-cov.txt''.
\item[\code{fnameirt}:] The name of the file containing additional information to link item parameters with
the covariance matrix and to identify parameters that have been constrained to a fixed value.
This is a file ending with ``-irt.txt''.
\item[\code{display}:] Logical value indicating whether the coefficients and the standard errors
should be printed. The default is \code{TRUE}.
\item[\code{digits}:] Integer value indicating the number of decimal digits to be used if display is \code{TRUE}.
\end{description}
The functions return a list with components:
\begin{description}
\item[\code{coef}:] The matrix with item parameter estimates.
\item[\code{var}:] The covariance matrix of item parameter estimates.
\end{description}
Item parameters are imported under the parameterization given in Equations~\ref{ltpar} and~\ref{lpar}.
The usual IRT parameterization can be obtained later by using function \code{modIRT}.
An example of importing data from the \pkg{ltm} package is given in Appendix~\ref{app}.
The \pkg{equateIRT} package does not handle mixed item types, but
this issue can be easily overcome by using the more general model and introducing constraints on some parameters.
For example, in the case of some item responses modeled as a two-parameter
logistic model and others as a three-parameter logistic model, the user can specify a three-parameter logistic
model for all items and fix the guessing parameter to zero for some items.
If item parameter estimates are obtained with IRT programs other than \pkg{flexMIRT}, \pkg{IRTPRO}, \pkg{ltm} or \pkg{mirt},
the user has to create a matrix containing the item parameter estimates. These can be provided using typical
IRT parameterization~(\ref{pr}) or with parameterization~(\ref{ltpar}) and/or~(\ref{lpar}).
In this matrix guessing, difficulty and discrimination parameters should strictly be given in this order and they are contained in different columns of the matrix. It is important that the names of the rows of the matrix
are the names of the items because this information is used to link different forms.
The covariance matrix is only necessary if the user is interested in obtaining the
standard errors of the equating coefficients.
It is important that the order of the items in the covariance matrix is the same as in the matrix with the
item parameter estimates. So, there are first guessing parameter, then difficulty parameters and
finally discrimination parameters.
An example of item parameters and covariance matrices organized in this way is given
in the datasets contained in the \pkg{equateIRT} package.
The package includes three simulated datasets for
illustrative purposes containing item parameter estimates
and covariance matrices of five forms. The item parameter estimates
were obtained with package \pkg{ltm}.
In particular,
dataset \code{est3pl} contains parameters of a three-parameter logistic model,
dataset \code{est2pl} contains parameters of a two-parameter logistic model and
dataset \code{estrasch} contains parameters of a Rasch model.
Each dataset is composed of a list of length 2 with components:
\begin{description}
\item[\code{coef}:] A list of length 5 containing the matrices
of item parameter estimates.
\item[\code{var}:] A list of length 5 containing the covariance
matrices of item parameter estimates.
\end{description}
The following section explains how to obtain the equating coefficients.
\subsection{Data analysis}
In this section, we use the dataset \code{est2pl} to illustrate the use of
the package. The code for loading the package and the data is
<<>>=
library("equateIRT")
data("est2pl", package = "equateIRT")
@
To perform the equating, it is necessary to reorganize the data using
function \code{modIRT}. This function creates an object of class `\code{modIRT}'
consisting of a list with length equal to the number of forms where each element
contains a lists with components:
\begin{description}
\item[\code{coefficients}:] Item parameter estimates.
\item[\code{var}:] Covariance matrix of item parameter estimates.
\item[\code{itmp}:] Number of item parameters of the model.
\end{description}
The names of the forms can be specified in function \code{modIRT}
using argument \code{names}. If \code{names} is not specified,
function \code{modIRT} assigns the names. If item parameters are
given under the latent trait parameterization, use option
\code{ltparam = TRUE}. If guessing parameters of a three-parameter
logistic model are given under the logistic parameterization, use
option \code{lparam = TRUE}. The default values of both these
arguments is \code{TRUE}, so the user does not need to specify them if
these parameterizations were used in the estimation of the parameters.
Using these options, the item parameters are returned under the usual
IRT parameterization as in Equation~\ref{pr}. In this case, the
covariance matrix of the item parameters under parameterization of
model~(\ref{pr}) is obtained, by using the delta method, on the basis
of the covariance matrix of item parameters under the
parameterizations ~(\ref{ltpar}) and/or~(\ref{lpar}), supplied by the
user. If item parameters already conform to the traditional
parameterization, the options \code{ltparam} and \code{lparam} should
be set to \code{FALSE}, and the function does not perform any
transformation. Argument \code{coef} is used to specify the item
parameter estimates. They should be given as a list of matrices (one
for each form) whose row names are the names of the items. Argument
\code{var} can be used to specify the covariance matrix of item
parameter estimates and it should be a list of matrices. If it is left
equal to \code{NULL}, standard errors of equating coefficients will
not be computed. Option \code{display = TRUE} can be used to print
item parameter estimates and the corresponding standard errors in
order to compare them with those provided by the software used to
estimate the item parameters. The coefficients can be extracted using
the \code{coef} method.
<<>>=
test <- paste("test", 1:5, sep = "")
mod2pl <- modIRT(coef = est2pl$coef, var = est2pl$var, names = test,
display = FALSE)
coef(mod2pl$test1)[1:5]
@
The linkage plan can be inspected using the function \code{linkp}.
Given a list of item parameter estimates with item labels as row names,
this function computes the number of common items between all pairs of forms
and returns a matrix whose elements indicate the number
of common items between the forms. On the diagonal of the matrix
there are the number of items of each form.
<<>>=
linkp(coef = est2pl$coef)
@
The output of the function shows that every form is composed of 20
items and presents 10 items in common with adjacent
forms. Furthermore, Forms~1 and 5 present 10 common items.
Function \code{direc} calculates direct equating coefficients between
two forms with common items. Arguments \code{mod1} and \code{mod2}
are objects of class `\code{modIRT}' containing item parameter coefficients
and their covariance matrix.
Argument \code{method} indicates the equating method to\linebreak use and should be one of
\code{"mean-mean"}, \code{"mean-gmean"}, \code{"mean-sigma"},
\code{"Haebara"} or\linebreak \code{"Stocking-Lord"}.
For example, to calculate direct equating coefficients between Form
1 and Form 5 using the Haebara method for the \code{est2pl} dataset
the code is
<<>>=
l15 <- direc(mod1 = mod2pl[1], mod2 = mod2pl[5], method = "Haebara")
l15
@
The \code{direc} function approximates the integrals in Equations~\ref{h}
and~\ref{sl} using a Gauss-Hermite quadrature with 30 points by default.
The number of quadrature points can be modified using argument \code{nq}.
Alternatively, setting \code{quadrature = FALSE} the integrals are replaced
with a sum over 40 equally spaced values ranging from $-4$ to 4 with an increment
of 0.05 and weights equal to one for all values.
Argument \code{D} can be used to specify the value of the constant
$D$ in model~(\ref{pr}) used in the estimation of item parameters.
A \code{summary} method for displaying the output of the function is available.
<<>>=
summary(l15)
@
The package provides also function \code{alldirec} to calculate direct
equating coefficients between all pairs of forms that present common items.
Using the mean-mean method, the code for our example is
<<>>=
direclist2pl <- alldirec(mods = mod2pl, method = "mean-mean")
direclist2pl
@
The \code{summary} function can be used to display all the equatings
performed
<>=
summary(direclist2pl)
@
or just one of them
<<>>=
summary(direclist2pl, "test1.test5")
@
In order to compute chain equating coefficients, it is necessary to
have previously computed direct equating coefficients using function
\code{alldirec}. Chain equating coefficients can be computed using
function \code{chainec}. This function requires the specification of
the length of the chain using argument \code{r}. For example, to
compute all chain equating coefficients of length 4 for the
\code{est2pl} dataset the code is
<<>>=
cec4 <- chainec(r = 4, direclist = direclist2pl)
cec4
@
The summary function displays all the equatings performed.
<>=
summary(cec4)
@
It is also possible to display only one equating using the following
code
<<>>=
summary(cec4, "test1.test2.test3.test4")
@
Option \code{f1} can be used to restrict the equatings to those chains
that start from the same form. To consider all chain equatings of
length 4 that start from Form 1 the code is
<<>>=
cec4.1 <- chainec(r = 4, direclist = direclist2pl, f1 = "test1")
@
Option \code{f2} can be used to restrict the equatings
to those chains that end with the same form.
To consider all chain equatings of length 4 that start from
Form 1 and end with Form 4 the code is
<<>>=
cec1234 <- chainec(r = 4, direclist = direclist2pl, f1 = "test1",
f2 = "test4")
@
Alternatively, it is possible to specify one or more particular paths
using argument \code{pths}. For example, to compute chain equating
coefficients between Forms~1 and 4 using path $\{1,5,4\}$ the code is
<<>>=
pth1 <- paste("test", c(1, 5, 4), sep = "")
pth1 <- data.frame(t(pth1), stringsAsFactors = FALSE)
cec154 <- chainec(direclist = direclist2pl, pths = pth1)
summary(cec154)
@
Another example is the chain equating of
Forms~1 and 5 using path $\{1,2,3,4,5\}$ .
<<>>=
pth2 <- paste("test", 1:5, sep = "")
pth2 <- data.frame(t(pth2), stringsAsFactors = FALSE)
cec12345 <- chainec(direclist = direclist2pl, pths = pth2)
summary(cec12345)
@
When two forms can be linked through different paths, the equating
coefficients can be averaged using function \code{bisectorec}, that
implements the bisector method. Options \code{weighted = TRUE} and
\code{unweighted = TRUE} can be used to calculate the weighted
bisector coefficients and the unweighted bisector
coefficients. Weights are determined in order to minimize
function~(\ref{of}). For example, to calculate bisector and weighted
bisector coefficients to equate Forms~1 and 4 through paths
$\{1,2,3,4\}$ and $\{1,5,4\}$, and Forms~1 and 5 through paths
$\{1,2,3,4,5\}$ and $\{1,5\}$ the code is
<<>>=
ecall <- c(cec1234, cec154, cec12345, direclist2pl["test1.test5"])
fec <- bisectorec(ecall = ecall, weighted = TRUE, unweighted = TRUE)
fec
@
The \code{summary} function displays the results.
<<>>=
summary(fec)
@
Function \code{eqc} can be used to extract a data frame containing
the equating coefficients.
For direct equating coefficients the code is
<<>>=
eqc(l15)
@
while, for a list of direct equating coefficients, the code is
<<>>=
eqc(direclist2pl)
@
It is also possible to select a specific link.
<<>>=
eqc(direclist2pl, link = "test1.test5")
@
For a list of chain equating coefficients it is possible to display the equating coefficients
for all paths, or specify a particular link and/or path.
The code is as follows
<<>>=
eqc(cec4)
eqc(cec4, path = "test1.test2.test3.test4")
@
Also for the output of function \code{bisectorec}, it is possible to extract a
data frame containing all the coefficients, or select the link and/or
the path.
<<>>=
eqc(fec)
eqc(fec, link = "test1.test4", path = "bisector")
@
Functions \code{direc}, \code{alldirec} and \code{chainec} provide
also the scale conversion of item parameter estimates. This is
included in the data frame \code{tab}, that is part of the outputs of
the functions. Function \code{itm} extracts this data frame. Some
examples for direct and chain equating coefficients are given below.
<<>>=
itm(l15)[1:3, ]
itm(direclist2pl, "test1.test5")[1:3, ]
itm(cec12345, "test1.test2.test3.test4.test5")[1:3, ]
@
The transformation of item parameter estimates can be found under
column \code{test1.as.test5}.
Since the \code{bisectorec} function returns the equating coefficients
of a plurality of scale conversions, it does not provide the scale
conversion of the item parameters. For average equating coefficients
the user can employ function \code{convert}, that is a specific
function to convert item and person parameters, given the equating
coefficients. The following code extracts the equating coefficients
obtained with the bisector method to transform from the scale of Form
1 to the scale of Form 4 and performs the scale conversion of item
parameters of Form 1 and of an hypothetical vector of person
parameters.
<<>>=
eqc14 <- eqc(fec, link = "test1.test4", path = "bisector")
convert(A = eqc14$A, B = eqc14$B, coef = coef(mod2pl$test1),
person.par = seq(-3, 3, 0.5))
@
Function \code{convert} can also be used for direct or chain equating parameters.
\section{Conclusion}
\label{con}
The linkage plans can be rather complex involving many forms, several
links, chains and the connection of forms through more than one path.
This situation necessitates the performance of chain equating in order
to link forms that do not present common items. Furthermore, when two
forms can be connected through different paths it can be useful to
synthesize the conversions obtained into a single one. The
\pkg{equateIRT} package includes not only the computation of direct
equating coefficients but also permits the computation of chain and
average equating coefficients. Another important and exclusive
feature of the \pkg{equateIRT} package is the provision of analytical
standard errors of equating coefficients. Standard errors are an
important tool for assessing the accuracy of the equating process and
can be also used to perform further inferential analysis.
\section*{Acknowledgments}
The author wishes to thank the referees and the Associate Editor for
their helpful comments and suggestions that greatly improved this
work.
\bibliography{equateIRT}
\newpage
\begin{appendix}
\section[Importing data from the ltm package]{Importing data from the \pkg{ltm} package}
\label{app}
This appendix shows how to import the estimates obtained with the
\pkg{ltm} package using the dataset \code{data2pl}. This dataset is
a list of length five and includes five simulated data frames
generated from a two-parameter logistic model. Each data frame
contains 5000 dichotomous responses to 20 items. Item parameter
estimates and covariance matrices included in the dataset
\code{est2pl} are obtained from the dataset \code{data2pl}.
The code for loading packages \pkg{ltm} and
\pkg{equateIRT} and the data, and for estimating a two-parameter logistic
model is
<<>>=
library("ltm")
library("equateIRT")
data("data2pl", package = "equateIRT")
m1 <- ltm(data2pl[[1]] ~ z1)
m2 <- ltm(data2pl[[2]] ~ z1)
m3 <- ltm(data2pl[[3]] ~ z1)
m4 <- ltm(data2pl[[4]] ~ z1)
m5 <- ltm(data2pl[[5]] ~ z1)
@
Function \code{import.ltm} extracts the item parameter estimates and
the covariance matrix.
<<>>=
estm1 <- import.ltm(m1, display = FALSE)
estm2 <- import.ltm(m2, display = FALSE)
estm3 <- import.ltm(m3, display = FALSE)
estm4 <- import.ltm(m4, display = FALSE)
estm5 <- import.ltm(m5, display = FALSE)
estm1$coef[1:3, ]
estm1$var[1:3,1:3]
@
The item parameters and the corresponding covariance matrix are given
under parameterizations~(\ref{ltpar}) and~(\ref{lpar}). The function
\code{modIRT} transforms item parameters into the usual IRT
parameterization given in Equation~\ref{pr} and returns item
parameter estimates and the covariance matrices in a format suitable
for running the equating.
<<>>=
estc <- list(estm1$coef, estm2$coef, estm3$coef, estm4$coef, estm5$coef)
estv <- list(estm1$var, estm2$var, estm3$var, estm4$var, estm5$var)
mod2pl.ltm <- modIRT(coef = estc, var = estv, display = FALSE)
@
\end{appendix}
\end{document}