The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.
pvRegression() now returns an object of class
"pvRegression" (still a data frame) with
print(), summary(), coef(),
vcov(), confint(), predict() and
nobs() methods. vcov() gives the full
Rubin-combined covariance matrix of the coefficients.pvMeanVar() has been removed. Use pvMean()
and pvVar() (or pvDescribe()) instead, which
also support weights and replicate standard errors.genResponses() gains a debug argument.
When debug = TRUE it returns a list holding the responses
and the intermediate data from each step: the standardised inputs, the
parsed perturbations, the perturbed item parameters, which perturbation
was applied to each case and item, the random draws, and the category
probabilities used (and those under the unperturbed items). Responses
are identical to debug = FALSE for the same seed.genResponsesProbs() tabulates the response
probabilities used for each case, item, and category alongside the
unperturbed probabilities, to check that perturbations were applied to
the intended cases. See the “Generate item responses” vignette.genResponses(perturbR = ) can now
apply to all cases (omit group, or use NULL or
NA), one group, or several groups (e.g.,
group = c("g1", "g2")). Elements can be given by name.pivotAt = "item",
which changes the discrimination around a point pivot
logits from the item’s location, so that the item location does not move
(category probabilities are unchanged at the pivot point). The default,
pivotAt = "theta", keeps the previous behaviour. A separate
shift can be added to move the item location without moving
the discrimination pivot.repMethod = "Fay",
with fayK), and the delete-one ("JK1") and
paired ("JK2") jackknife. All support a grouping variable
(by). See the “Population statistics from plausible values”
section of the introductory vignette.
pvStat() is the core function: mean, variance, SD,
median, quantiles, IQR, sum of weights, or a user-supplied
function(x, w).pvMean(), pvVar(), pvSd(),
pvMedian(), pvQuantile(), pvIqr()
and pvDescribe() are wrappers for common descriptive
statistics.pvDiff() estimates differences between groups (e.g.,
girls minus boys), computing the difference within each PV and replicate
so the standard error reflects covariance induced by the sampling
design.pvCor() estimates correlations between a measure and
one or more other measures, combined on the r scale (default) or the
Fisher-z scale (combine = "fisherZ"), with confidence
intervals. Each measure may be latent (a set of PVs) or manifest (a
single observed column).pvCorMatrix() and pvCovMatrix() estimate
the correlation and covariance matrices of several measures, e.g., the
outcome domains of IELS, PISA or TIMSS. Latent and manifest measures can
be mixed: latent measures are paired by PV draw, and a manifest
variable’s column is reused for every draw. With only manifest
variables, the ordinary weighted matrix is returned with replicate
standard errors. Results are available in long or matrix form. Cases are
dropped listwise so the matrix is valid; a warning is given if Fisher-z
combining produces a matrix that is not positive semi-definite.pvRegression() fits a weighted linear regression of PVs
on background variables given as a one-sided formula, and returns
coefficients and R-squared with standard errors and confidence
intervals.pvMeanVar() is now a thin wrapper around
pvMean() and pvVar(). Results are unchanged.
Its documentation now correctly describes the input as n cases by m
PVs.genResponses() has been split into steps
(genRespInputs(), parsePerturbations(),
perturbItem(), buildPerturbMap(), and
genMissingMCAR()), each of which can be called on its
own.genResponses() now gives informative errors for invalid
perturbations (e.g., a missing factor or
pivot, or a group not found in
groups) and for mcarP outside 0 to 1.pivot in
genResponses() has been corrected: it is added to the delta
dot. For a dichotomous item, perturbed and unperturbed curves cross at
theta = pivot / (a * (factor - 1)).getCqParams() (and so getCqRespModel())
removed the wrong parameters when a model had unidentified (problem)
gins, misaligning parameter estimates with their gins and, in some
cases, failing with “arguments imply differing number of rows”.ReadSys and the Read* helpers) have been
rewritten in C++. The file is read into memory once and parsed in C++;
ConQuestSys() is roughly 10-20 times faster on the included
examples, and the results are identical.Read* helpers now take a ‘cqReader’ (an
in-memory file created internally by ConQuestSys() and
ConQuestRout()) rather than a connection.
DecompressSys() now takes and returns a raw vector, and no
longer writes a temporary file.conquestr::lcs and
conquestr::lcs_vecconquestr::getCqItanal, and
conquestr::getCqItanalSummary for example when group
variables or levels within groups contain punctuation (e.g.,
“my-group_variable.”, and “my-group_level_1”)conquestr::getCqItanal, and
conquestr::getCqItanalSummary)returnObject[[1]][[4]]
instead of returnObject[[4]]?conquestr::getCqHistory.plotCCCplotModelICCgetCqRespModelConQuestCall now will call “cmd.exe” via
system2 on Windows. This is morre consistent with behaviour
on Mac.ConQuestCall with paths with spaces. Fixed.ConQuestCall now will try and find the ConQuest
executable rather than requiring explicit declarationplotRout to support plotting
rout obejctsplotRout has methods for ICC and for Infomap
objects.searchConQuestSys to help users search names in
system file objectsreplaceInDataFrame replaces
zapSystemMissing in calls to createDfFromSys -
this uses Rccp in place of base R and is approx 25 times faster making
this function usable for large analyses.These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.