NEWS

pwrss 1.3.3

Added \donottest around some examples to prevent the package failing at the pretests on CRAN (it requires that examples run through within 5 secs)

Fixed a mistake in the calculation of the density function for lambda prime, implemented more rigorous unit tests for the lambdap-functions (comparing against the R-packages MBESS and effectsize)
Adjustments to cater for noLD being added as additional test in the CRAN package checks, and extended GitHub action R-CMD-check to include OS / R combinations recently added as additional test in the CRAN package checks
Smaller improvements to the vignette and inst/CITATION (takes date / year and version from the meta-data)
Smaller improvements to satisfy the requirements / suggestions by the goodpractice R-package

Replaced functions to calculate lambda prime (formerly provided by the R-package sadists) with own implementations (numerically, these implementations reveal results close to those provided by sadists; however, a problem with a numerical instability in that package doesn't exists for these implementations, and df's can be estimated in power.lp.test)
Smaller improvements to satisfy the requirements / suggestions by the goodpractice R-package
Fix a bug arising from stats::poly() returning very small values (~1e-16) in places where 0 is expected
Small improvement to the unit tests, preventing that snapshots are deleted if a comparison is not run

Added unit tests for all procedures
Added minimum detectable effects
Added exact correlation (one sample)
Added type S and M errors for two-sided tests, available through plot() function
Added clean.parms, unifying how the input parameters in the results list are returned
Added further input checks so that all input parameters are checked
Simplified code in power.z.logistic and power.z.poisson
Corrected formula for Hsieh et al. (1998) in power.z.logistic
Added target.effect to determine effect of interest in power.f.ancova

Addition 1: ANCOVA procedures with means and contrasts (Shieh)
Addition 2: Exact tests for proportions (Fisher & McNemar)
Addition 3: Generic binomial test (with plotting feature)
Addition 4: Dependent correlations (Steiger)
Addition 5: Effect size conversion functions
Extension 1: Generic tests handles non-zero null (useful for equivalence and minimum effect testing). Some specific tests also allow non-zero null.
Extension 2: ANCOVA procedures allow R-squared adjustment (i.e., Keppel and Shieh)
Consistent function names, population-consistent argument names (Greek notation) and base R consistency
Rigorous error checks and warnings
Comprehensive print utility with three-tier verbosity control
Minor bug fixes to plot() function

alternative = "less" now produce correct power rates in pwrss.z.prop() function. Thanks to Leszek Gawarecki for reporting the issue.
The default for p0 argument in pwrss.chisq.gofit() can now be overwritten. Thanks to Kate Crespi for reporting the issue.
Type 1 and type 2 error plots show and print correct power estimates for two-tailed tests when effect size is near zero. Thanks to dpnichols811 (GitHub profile handle) and Adrian Olszewski for reporting the issue.
pwrss.z.corr() print correct errors with alternative = "greater" when in fact it is less (or vice versa). Thanks to Jarrod Hadfield for reporting the issue.

Added pwrss.np.2means(), pwrss.z.logreg(), pwrss.z.poisreg(), pwrss.chisq.gofit(), and generic power.chisq.test() functions
Improvements to plot() function
Improvements to documentation
Improvements to pwrss.z.med() and pwrss.f.ancova() console print
More detailed and comprehensive examples in vignette
Bug fixes (Welch’s t test)