Changelog • PROsetta

PROsetta 0.4.2

Fixed Rd cross-reference NOTE by adding package anchors to \link{} and \linkS4class{} targets.
Updated documentation for runEquateObserved() and SingleGroupClass to ensure proper linking to external packages.

Added support for working with GPC models.
Added support for when items between instruments have different item models. (e.g., the anchoring instrument uses GR, and the instrument being calibrated uses GPC.)
Added support for when items within the anchoring instrument use different item models. (e.g., the anchoring instrument uses GR for some items and GPC for others.)
Added support for when items within the instrument being calibrated use different item models. (e.g., the instrument being calibrated uses GR for some items and GPC for others.)

Fixed where loadData() would not warn about items that may need reverse coding for scale IDs other than 1.

Now requires TestDesign (>= 1.5.1).
Response probabilities are now computed faster using cpp functions.
Deprecated guiPROsetta() is removed. Use PROsetta() instead.
Updated documentation.
Made minor updates to the Shiny app in PROsetta().
Removed ncat column in anchor parameters. This is now inferred from the number of parameters.
Added cpp routine for EAP computation after Lord-Wingersky recursion. This improves the speed of runRSSS().
getRSSS() now nudges the user if the prior mean input looks like a T-score, which should be entered in the theta metric.
Updated CITATION to use bibentry() to meet CRAN requirements.

runRSSS() output now includes linear approximation betas when the CPLA method is used.
runLinking() output now includes the latent mean and variance when the FIXEDPAR method is used.
runCalibration(), runLinking(), runEquateObserved(), getCompleteData() gains verbose argument for printing status messages. Status messages that were used to be printed in previous versions are now suppressed by default.

Removed unused columns (min_score, reverse, scores) in example datasets for clarity. The package functions do not use these columns.
loadData() now warns if there is a variable that may need reverse coding. This is triggered by a negative correlation value.

Fixed where runLinking(method = "FIXEDPAR") was not working when the anchor instrument ID was not 1 in item map.
Fixed where runLinking(method = "FIXEDPAR") was not working when the anchor and target instruments had different numbers of categories in response data.
Fixed where runFrequency() was not sorting categories correctly when the number of categories was 10 or above.
Fixed where runCalibration(fixedpar = TRUE) was not reading anchor parameters correctly when an integer value existed in anchor parameters.
Fixed where item parameters for dichotomous items were triggering an error while being parsed.
For compatibility with R < 4.0, loadData() now sanitizes input data when a data frame is supplied.

runLinking() now supports method = 'CPFIXEDDIM' to perform two-dimensional calibration, for use in performing calibrated projection (Thissen et al., 2015). The difference with method = 'CP' is that 'CPFIXEDDIM' constrains the mean and the variance of the latent anchor dimension, instead of constraining anchor item parameters. For this purpose, a unidimensional fixed parameter calibration using only the anchor response data is performed to obtain the mean and the variance.
getRSSS() for computing a single raw-score to standard-score table is now exposed.

Added getResponse() for extracting scale-wise response data from a PROsetta_data object.
Added getItemNames() for extracting scale-wise item names from a PROsetta_data object.

runLinking() now supports method = 'CP' to perform two-dimensional calibration, for use in performing calibrated projection (Thissen et al., 2011).
runLinking() now supports method = 'CPLA' to perform two-dimensional calibration, for use in performing linear approximation of calibrated projection (Thissen et al., 2015).
runRSSS() now performs two-dimensional Lord-Wingersky recursion with numerical integration, when the output from runLinking(method = 'CP') is supplied.
runRSSS() now performs linear approximation of calibrated projection, when the output from runLinking(method = 'CPLA') is supplied.
Shiny application PROsetta() now supports calibrated projection and its linear approximation.

PROsetta_config class and createConfig() are now deprecated. The functionalities are merged to PROsetta_data class and loadData().
run*() functions now require PROsetta_data objects instead of PROsetta_config objects.
runLinking() now has method argument to specify the type of linking to perform. Accepts MM, MS, HB, SL, and FIXEDPAR.

runLinking() is now capable of performing fixed calibration.
runCalibration() now performs free calibration by default.
runCalibration() and runLinking() now errors when iteration limit is reached, without returning results.
runRSSS() now returns thetas in addition to T-scores, and also expected scores in each scale.
runEquateObserved() now has type_to argument to specify direct raw -> T-score equating or regular raw -> raw equating.
Functions are now more verbose.
Added PROMIS Depression - CES-D linking dataset data_dep.
Added plot() for drawing raw score distribution histograms.
Added plotInfo() for drawing scale information plots.
Added several helper functions.
Made cosmetic improvements on Shiny app.

Fixed where Shiny app was displaying SL method linear transformation constants regardless of specified linking method.

Added scalewise argument to runClassical() and runCFA(). When TRUE, analysis is performed for each scale.
runEquateObserved() now removes missing values to produce correct raw sums.
loadData() now retains missing values.