Skip to content

dscore 1.9.0

Latest
Latest
Compare
Choose a tag to compare
@stefvanbuuren stefvanbuuren released this 12 Jun 21:45
· 36 commits to master since this release
v1.9.0
acb1b52

Overview

This is a major update of the dscore package featuring:

  • a new default reference "preliminary_standards"
  • a correction of an issue with the scaling factor
  • a major clean-up of the itembank, references, keys, and R code
  • improved documentation and examples

Major issues

BREAKING CHANGE: On May 31, 2024 we detected a long-time error in the calculation of the D-score resulting from an incorrect scale factor that led us to believe that item characteristic curves are steeper than the actually are. The impact of the error on the result is as follows:

  1. There is no effect on the difficulty estimates of the Rasch models,
  2. The D-score estimates are slightly altered but changes are small,
  3. The references are largely unaffected and need not to be redone,
  4. The estimates of the SEMs can differ substantially, so inferences based on the SEMs should be re-evaluated,
  5. When there were changes in the analyses, the results in the newer method look smoother and are preferred.

The correction appeared in the development version dscore 1.8.8, and is now incorporated into release dscore 1.9.0. For backward compatibility to dscore 1.8.7 and earlier, use the argument algorithm = "1.8.7" in calls to the dscore() function.

BREAKING CHANGE: dscore_posterior() now returns a data.frame with column names that indicate the quadrature point. This was an unnamed matrix. Code that expects a matrix as the return of dscore_posterior() may need to be adapted.

NEW DEFAULT KEY: Adds a new reference "preliminary_standards" calculated from selected subsample of the GSED Phase 1 data, and makes these the default in this release. The reference is a temporary stand-in for a future norm-based standard for normal early child development. This reference replaces the temporary reference "phase1_healthy" that was introduced in dscore 1.8.7. Compared to the "phase1" reference, the "preliminary_standards" reference has the following differences: 1) D-score estimation uses the new model 20240601 with correct scale factor, 2) Calculates the D-score for SF and LF separately (not combined), 3) Tunes the GAMSLSS model to fit the healthy subsample. The "phase1_healthy" object is removed.

Major changes

  • Adds a new age-conditional reference for population "dutch" calculated using the "gsed2212" key.
  • Defines a new key "gsed2406" to accomodate for the changed prior mean because of the adoption of the new reference "preliminary_standards" as the base population. The key "gsed2406" is identical to "gsed2212", and is the default key in this release.
  • Adds a new builtin_keys table that contains proper defaults for the base reference, transformation and quadrature points per key
  • Indexed a reference now by two fields: key and population. Previously the index was based on only population. This change allows for multiple references per key, and for references created for the same population under different keys. The key field is now mandatory in the reference table.
  • Adds a verbose option to dscore(), dscore_posterior(), get_age_equivalent(), get_reference(), get_tau(), daz() and zad() to print progress messages to the console on the values of key, population, transform, qp and algorithm. This is useful for debugging and for understanding the behavior of the functions.
  • Cleans up the R code to take advantage of the specification made in the new builtin_keys table. This makes the code more readable and maintainable.
  • Retires keys sf2206, lf2206, 294_0, gsed2206, gsed2208 and removes from the builtin_itembank.
  • dscore() and dscore_posterior() can now copy variables from the input data into the output through the prepend argument. (#46)

Minor changes

  • Simplifies the package DESCRIPTION file
  • New internal init_key() and set_default_xxx() functions to regulate values for key, population, transform and qp arguments
  • Renames the files in data-raw/data/keys to more consistent names, adapts data-raw/R/save_builtin_itembank.R to reflect model history, and rebuilds builtin_itembank
  • Removes the dependency on tibble and tidyselect, and replaces the dependency on stringr by the lighter stringi
  • Rename the argument name reference to reference_table in daz() and zad() to avoid confusion with the references argument in get_reference()
  • Simplifies the spelling of the term "D-score" to improve consistency and readability
  • Replaces magrittr pipe %>% by base pipe |>
  • Make style more consistent with styler
  • Per request from CRAN (Specified C++11: please drop specification unless essential), removes a C++11 specification
  • Moves error evasion code into internal pBCT()
  • Document up-rounding to a D-score of 1 or higher when daz() and zad() using the BCT transformation for positive values
  • Removes the superfluous names attribute from the return value of daz() and zad()
  • Evades an error produced by internal pBCT() when is.na(nu) is TRUE
  • Renames GSED HH to GSED HF
  • Change CITATION file to use the bibtex package
  • Moved all keys to the data-raw/data/keys folder and renamed them to improve readability
Assets 2
Loading