Make sure that mob() from partykit
(rather than party) is called by all tree
functions.
Improved version of
node_profileplot(..., what = ...) which now allows to
visualize not only itempar() but also coef(),
threshpar and so on by setting what = "item"
or "coef" or "threshold" etc.
In previous versions the
gpcmtree(..., minsize = ...) argument was ignored. Now it
is passed on correctly to mob_control().
New functions npltree() and gpcmtree()
with corresponding visualization functions for model-based recursive
partitioning of parametric logistic IRT models (2PL, 3PL, 3PLu, 4PL,
Rasch/1PL) and the generalized partial credit model, respectively. The
output of these functions might still change slightly in future
versions.
Make sure that the “raschtree” vignette can also be compiled even
when the stablelearner package is not available.
stablelearner package (>= 0.1-3).IGNORE_RDIFF flags in some examples in order to
avoid showing diffs due to small numeric deviations in some checks
(especially on CRAN).The method behind mpttree() has now been published
by Behavior Research Methods, see doi:10.3758/s13428-017-0937-z.
The predict() method for bttree objects
works again correctly for type = "parameter" (previously
called type = "worth") and type = "rank".
Thanks to Heather Turner for pointing out the problem.
node_btplot() used to fail for
worth = FALSE or ref = <character> but
works now correctly again. Thanks to Heather Turner for pointing out the
problem. Also ref = <character> now either matches to
<object>$labels or (if not available) to
colnames(<coefficients>) to facilitate usage with the
PlackettLuce package.
node_regionplot() now also works correctly on
devices with a default white (rather than transparent) background such
as png() or jpeg().
node_* panel functions gained
bg = "white" argument to enable different background
filling color.
The recently added mpttree() function is now
accompanied by a working paper, see citation("psychotree")
or ?mpttree for details.
Properly imported grDevices in
NAMESPACE.
Basil Abou El-Komboz changed his name to Basil Komboz.
mpttree() and visualization function for
model-based recursive partitioning of multinomial processing tree (MPT)
models. These functions are somewhat experimental, and their user
interface might change in future releases.Internals of package psychotree completely
reorganized to employ the recent additions and changes in
psychotools (>= 0.3-0) and to use the new
partykit (>= 0.2-0) implementation of the
mob() function (rather than the old party
implementation).
The handling of argument ref when producing a
region/effect plot was changed. Whereas in the previous implementation,
the restriction specified in this argument was applied to the cumulative
absolute item threshold parameters, it now is applied to the absolute
item threshold parameters.
New panel-generating visualization functions
node_profileplot() and node_regionplot() have
been added which replace the old functions node_raschplot()
and node_effects().
pctree() and
rstree(), respectively. These require package
psychotools >= 0.2-0 which provides the corresponding
basic model fitting tools.DESCRIPTION.Adjusted vignette “raschtree” according to the slightly different
results due to the new approximation formulae for p values in the supLM
test used in party::mob() (>= 1.0.3).
Improved the package’s tests by adding reference output for the examples and vignette for comparison checks.
Moved all base functionality to the new psychotools
package in order to provide common infrastructure for both
psychotree and psychomix as well as other
packages. This affects the paircomp class and associated
methods as well as the model fitters btReg.fit() and
RaschModel.fit() and associated methods.
The data sets Firstnames,
GermanParties2009, and Soundquality have also
been moved to psychotools.
Enhanced raschtree() functionality, introduced in
detail in the new
vignette("raschtree", package = "psychotree").
Added subsample from SPIEGEL Studentenpisa (SPISA)
quiz for students from Bavaria that answered the same set of
questions.
Added artificial data set DIFSim exhibiting
differential item functioning, employed for illustrating Rasch
trees.
Added raschtree() function, based on simple
RaschModel() (for binary 0/1 items).
Some bug fixes (subsets with missing drop = FALSE)
for paircomp() and bttree() with only two
objects
First CRAN release of the package accompanying the forthcoming paper Accounting for Individual Differences in Bradley-Terry Models by Means of Recursive Partitioning by Strobl, Wickelmaier, and Zeileis, accepted for publication in Journal of Educational and Behavioral Statistics.
Compared to earlier package versions available from R-Forge, this
package does not depend on data/code from the prefmod2
package. All required functionality (basic Bradley-Terry regression,
paircomp class, data sets) are now contained in the
psychotree package.