- Added 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.
- More detailed vignette (typo fixes)
- Added web application links (shiny dashboards)
- Minor bug fixes with
plot() function
- 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)
- Added
pwrss.t.reg() and pwrss.z.med()
functions