Shiny app spit and polish

This is minor release that includes small bug-fixes in package functions and in the built-in shiny app.

• In the SCD_effect_sizes Shiny app:
  • Corrected a bug in the single-series calculator that caused an error when showing the graphed data.
  • Added a message in the single-series calculator to clarify minimum number of data points.
  • In the multiple-series calculator, redesigned the plotting interface the available choices for each filtering variable are calculated after applying the previous filters.

It's the little things

This is minor release that includes small bug-fixes and tweaks to package functions and the built-in shiny app.

• The package and SCD_effect_sizes app now accept uppercase outcome measurement scales for LRRi, LRRd, or LOR effect sizes.
• The package and SCD_effect_sizes app now accept uppercase labels for the direction of improvement variable.
• Corrected a bug in the single-series calculator that required data to be between 0 and 100 for the PoGO effect size.
• Updates to the multiple-series calculator in the SCD_effect_sizes app:
  • The app now shows a warning message if the improvement direction variable includes a category that is not acceptable.
  • The app now shows a warning message if a measurement scale is not acceptable for LRRi, LRRd, or LOR effect sizes.

PoGO

SingleCaseES 0.7.1

• Fixed a bug in the SCD_effect_sizes app (multiple-series calculator) which led to an error when calculating PoGO from an uploaded .xlsx file.
• Added several example datasets to the package.

SingleCaseES 0.7.0

• Fixed a bug in the SCD_effect_sizes app (multiple-series calculator), which led to incorrect R syntax when phase_pair_calculated is included as a grouping variable.
• Added function for calculating Percent of Goal Obtained effect size estimate.
• Updates to SCD_effect_sizes() shiny app.
  • Added Percent of Goal Obtained effect size measure to single-series calculator and multiple-series calculator.
• Changes to internal functions because of syntax changes in dplyr 1.1.0.

Updated documentation

This is minor release that includes new features in the built-in shiny app and updates to the documentation and vignettes.

• Revised in-app documentation for the SCD_effect_sizes web app. Added links to video demonstrations.
• Changes to the SCD_effect_sizes shiny app
  • The multiple series calculator gains an option for changing the number of digits displayed in the output.
  • Fixed a bug in the SCD_effect_sizes app (multiple-series calculator) plotting facilities, which caused an error if the phase variable was numeric.
  • Fixed a bug in the multiple-series calculator so that bintervals, bobslength, and blrrfloor are used for calculating LOR and LRR effect sizes.
  • Added an option to calculate phase pair numbers in the multiple-series calculator for uploaded data sets.
• Added informative error messages to the truncation constant calculations.

Weighting options and graphs

This is a major release that includes new options for batch_calc_ES(), including an expanded selection of weighting schemes for use when aggregating effect size estimates, and several new features for the SCD_effect_sizes shiny app, including an interface for taking a subset of observations before conducting batch calculations and an interface for visualizing the data used for batch calculations. This release also includes some tweaks for specific effect size functions, including reporting the SD used for standardization in calculating within-case standardized mean differences with SMD().

• Changes to batch_calc_ES()
  • batch_calc_ES() now provides several additional options for weights to use when aggregating effect sizes, including weighting by nA (the number of baseline phase observations), nB (the number of treatment phase observations), nA * nB, or 1 / nA + 1 / nB.
  • For aggregating effect sizes, batch_calc_ES() now defaults to equal-weighting (rather than inverse-variance weighting)
  • For aggregating effect sizes, batch_calc_ES() now provides more informative error messages for effect sizes that do not have standard errors (and therefore cannot be aggregated with inverse-variance weights).
  • Corrected a cosmetic bug in a warning message of batch_calc_ES() when scale argument is specified as a variable and ES includes "LOR".
• Changes to functions for specific effect sizes
  • Added intervention_phase argument to NAP(), SMD(), and other user-facing effect size calculation functions.
  • The SMD() function now reports the SD (baseline SD or pooled SD) used in the denominator of the standardized mean difference.
  • The SMD() function now returns a warning instead of an error when the standard deviation is zero.
  • Added option for Kendall's rank correlation in the Tau_BC() function.
  • Added warn argument to Tau_BC() function, to control printing of messages regarding baseline trend test.
• Changes to the SCD_effect_sizes shiny app
  • The batch entry calculator now includes an interface for visualizing data series.
  • The batch entry calculator now includes an interface for taking a subset of observations.
  • Miscellaneous aesthetic and layout fixes.
  • Added unit tests of shiny app results.

New effect sizes, ES aggregation, and replication code

This is major release that includes new effect size calculation functions, new post-processing features, and updates to the built-in shiny app. Changes were as follows:

• batch_calc_ES() gains an argument aggregate, for post-processing effect size estimates by averaging across the levels of a variable.
• Added functions for calculating two new effect sizes:
  • log ratio of medians
  • baseline-corrected Tau
• Updates to SCD_effect_sizes() shiny app.
  • Added aggregate feature.
  • Added log ratio of medians and baseline-corrected Tau effect size measures.
  • Allowed use of .xlsx files.
  • R code for replicating the batch calculation.

Compatibility with tidyr 1.0.0

This release includes updates to internal functions for compatibility with tidyr 1.0.0 and user interface improvements for an embedded shiny app.

• Updated internal functions for compatibility with tidyr 1.0.0.
• Tweaks for SCD_effect_sizes() shiny app.
  • Corrected bug in the single-series calculator that ignored direction of improvement for parametric effect sizes.
  • Fixed bug in the batch calculator phase-selection drop-down menus.
  • Corrected bug in the batch calculator so that confidence intervals other than 95% could be calculated.
  • Added a black-box warning any time PND is calculated.
  • Added option to convert LRRi and LRRd to percentage change.

Shinier shiny app

This release includes some improvements to the user interface of the SCD_effect_sizes() shiny app, along with miscellaneous bug fixes.

• Added check for purrrlyr package when loading gem_scd calculator.
• Corrected bug in degrees of freedom for within-case standardized mean difference with baseline SD.
• Tweaks for SCD_effect_sizes() shiny app.
  • Improved UI for batch calculation mode.
  • Added the option to select a treatment phase for batch calculation mode.
  • Allowed for choice between baseline SD and pooled SD for the within-case standardized mean difference. Previously, the option was available but not actually used for calculation.

Compatibility with dplyr 0.8.0

This is a maintenance release, which updates internals for compatibility with upcoming release of dplyr 0.8.0.

• Fixed bug in SCD_effect_sizes() shiny app that caused an error in batch calculator when LRRi, LRRd, or LOR were not selected.
• calc_ES() now returns a zero-length data frame if condition and outcome arguments are both length-zero vectors.
• calc_ES() errors if condition and outcome arguments are not the same length.

First CRAN release

This is the first version of the package released to CRAN. It includes many new features:

• The calc_ES() function for calculating multiple effect size estimates on a single data series.
• The batch_calc_ES() function for calculating effect size estimates for multiple data series.
• The gem_scd() function for estimating the gradual effects model (Swan & Pustejovsky, 2018).
• A shiny app for calculating basic within-case effect sizes.
• A shiny app for estimating effect sizes based on gradual effects model.
• Revised documentation.
• Extensive unit testing.