Structural Change Tests in Parametric Models
Performs model-based tests for structural change (or parameter instability) in parametric models.
## Default S3 method: sctest(x, order.by = NULL, functional = maxBB, vcov = NULL, scores = estfun, decorrelate = TRUE, sandwich = TRUE, parm = NULL, plot = FALSE, from = 0.1, to = NULL, nobs = NULL, nrep = 50000, width = 0.15, xlab = NULL, ...)
x |
a model object. The model class can in principle be arbitrary
but needs to provide suitable methods for extracting the |
order.by |
either a vector |
functional |
either a character specification of the functional
to be used or an |
vcov |
a function to extract the covariance matrix
for the coefficients of the fitted model:
|
scores |
a function which extracts the scores or estimating
function from the fitted object: |
decorrelate |
logical. Should the process be decorrelated? |
sandwich |
logical. Is the function |
parm |
integer or character specifying the component of the estimating functions which should be used (by default all components are used). |
plot |
logical. Should the result of the test also be visualized? |
from, to |
numeric. In case the |
nobs, nrep |
numeric. In case the |
width |
numeric. In case the |
xlab, ... |
graphical parameters passed to the plot method (in case
|
sctest.default is a convenience interface to gefp for
structural change tests (or parameter instability tests) in general
parametric models. It proceeds in the following steps:
The generalized empirical fluctuation process (or score-based CUSUM process)
is computed via scus <- gefp(x, fit = NULL, ...) where ...
comprises the arguments order.by, vcov, scores, decorrelate,
sandwich, parm that are simply passed on to gefp.
The empirical fluctuation process is visualized (if plot = TRUE) via
plot(scus, functional = functional, ...).
The empirical fluctuation is assessed by the corresponding significance test
via sctest(scus, functional = functional).
The main motivation for prociding the convenience interface is that these three steps can be easily carried out in one go along with a two convenience options:
By default, the covariance is computed by an outer-product of gradients
estimator just as in gefp. This is always available based on the scores.
Additionally, by setting vcov = "info", the corresponding information
matrix can be used. Then the average information is assumed to be provided by
the vcov method for the model class. (Note that this is only sensible
for models estimated by maximum likelihood.)
Instead of providing the functional by an efpFunctional
object, the test labels employed by Merkle and Zeileis (2013) and Merkle, Fan,
and Zeileis (2013) can be used for convenience. Namely, for continuous numeric
orderings, the following functionals are available:
functional = "DM" or "dmax" provides the double-maximum test (maxBB).
"CvM" is the Cramer-von Mises functional meanL2BB.
"supLM" or equivalently "maxLM" is Andrews' supLM test
(supLM). "MOSUM" or "maxMOSUM" is the MOSUM
functional (maxMOSUM), and "range" is the range
functional rangeBB. Furthermore, several functionals suitable
for (ordered) categorical order.by variables are provided:
"LMuo" is the unordered LM test (catL2BB),
"WDMo" is the weighted double-maximum test for ordered variables
(ordwmax), and "maxLMo" is the maxLM test for
ordered variables (ordL2BB).
The theoretical model class is introduced in Zeileis and Hornik (2007) with a
unifying view in Zeileis (2005), especially from an econometric perspective.
Zeileis (2006) introduces the underling computational tools gefp and
efpFunctional.
Merkle and Zeileis (2013) discuss the methods in the context of measurement invariance which is particularly relevant to psychometric models for cross section data. Merkle, Fan, and Zeileis (2014) extend the results to ordered categorical variables.
Zeileis, Shah, and Patnaik (2013) provide a unifying discussion in the context of time series methods, specifically in financial econometrics.
An object of class "htest" containing:
statistic |
the test statistic, |
p.value |
the corresponding p value, |
method |
a character string with the method used, |
data.name |
a character string with the data name. |
Merkle E.C., Zeileis A. (2013), Tests of Measurement Invariance without Subgroups: A Generalization of Classical Methods. Psychometrika, 78(1), 59–82. doi:10.1007/S11336-012-9302-4
Merkle E.C., Fan J., Zeileis A. (2014), Testing for Measurement Invariance with Respect to an Ordinal Variable. Psychometrika, 79(4), 569–584. doi:10.1007/S11336-013-9376-7.
Zeileis A. (2005), A Unified Approach to Structural Change Tests Based on ML Scores, F Statistics, and OLS Residuals. Econometric Reviews, 24, 445–466. doi:10.1080/07474930500406053.
Zeileis A. (2006), Implementing a Class of Structural Change Tests: An Econometric Computing Approach. Computational Statistics & Data Analysis, 50, 2987–3008. doi:10.1016/j.csda.2005.07.001.
Zeileis A., Hornik K. (2007), Generalized M-Fluctuation Tests for Parameter Instability, Statistica Neerlandica, 61, 488–508. doi:10.1111/j.1467-9574.2007.00371.x.
Zeileis A., Shah A., Patnaik I. (2010), Testing, Monitoring, and Dating Structural Changes in Exchange Rate Regimes, Computational Statistics and Data Analysis, 54(6), 1696–1706. doi:10.1016/j.csda.2009.12.005.
## Zeileis and Hornik (2007), Section 5.3, Figure 6
data("Grossarl")
m <- glm(cbind(illegitimate, legitimate) ~ 1, family = binomial, data = Grossarl,
subset = time(fraction) <= 1800)
sctest(m, order.by = 1700:1800, functional = "CvM")Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.