-
Due to a bug in previous versions of
WRS2::lincon
, even if p-value correction wasmethod = "none"
, it still applied Hochberg's correction and the{pairwiseComparisons}
package once again applied Holm's correction, which means the p-values were over-corrected for multiple comparisons. This bug has been fixed inWRS2 1.1-3
and therefore the users should expect slightly different p-values for between-subjects post hoc robust tests. -
{pairwiseComparisons}
now relies on{statsExpressions}
for statistical analysis. -
All included datasets have now been removed since the same datasets are also present in
{statsExpressions}
package. -
No longer depends on
ipmisc
package.
-
Maintenance and internal changes.
-
Improvements to docs.
-
To avoid confusion among users, the trimming level for all functions is now changed from
tr = 0.1
totr = 0.2
(which is whatWRS2
defaults to). -
The
...
are now passed to other methods. This can be used to specify additional arguments, likealternative
(#28). -
Gets rid of
iris_long
dataset, which was not used in the package.
-
Minor internal refactoring.
-
Removes
insight
from dependencies.
- Minor internal refactoring.
-
Minor internal refactoring.
-
Removes the unnecessary (and confusing)
significance
column from all outputs.
-
To be consistent with the rest of the
ggstatsverse
, the Bayes Factor results are now always shown in favor of null over alternative (BF01
). -
pairwise_comparisons
function getssubject.id
argument relevant for repeated measures design.
-
The
label
column returned inpairwise_comparisons
now displays the p-value adjustment method in the label itself. -
pairwise_caption
function has changed its output to reflect changes made to the p-value labels. -
Major internal refactoring to get rid of the following dependencies:
broomExtra
,dunn.test
,forcats
, andtidyr
. This comes at the cost of omission of few of the details that were previously included in the output (e.g.,mean.difference
column for Student's t-test).
- Hotfix release to fix failing tests due to release of
tidyBF 0.3.0
.
-
Fixes a bug which affected results for within-subjects design when the dataframe wasn't sorted by
x
(#19). -
This fix also now makes the results more consistent, such that irrespective of which type of statistics is chosen the
group1
andgroup2
columns are in identical order.
- Hot fix release to address failing tests on the old release of
R
(3.6
).
-
For repeated measures datasets with
NA
s present, the Bayes Factor values were incorrect. This is fixed. -
Internal refactoring to improve data wrangling using
ipmisc
.
-
Removes dependence on
jmv
and instead relies ondunn.test
andPMCMRplus
. This significantly reduces number of dependencies. -
The non-parametric Dwass test has been changed to Dunn test.
-
Adapts to breaking changes in upcoming release of
broom 0.7.0
. -
Thanks to Sarah, the package has a hexsticker. :)
-
Due to changes made to downstream dependencies, the minimum R version expected is bumped to
3.6.0
. -
Adds support for the Bayes Factor tests.
-
Exports the internal helper function
pairwise_caption
.
- Maintenance release to import functions from
ipmisc
.
pairwise_comparisons_caption
is removed since it was helpful only forggstatsplot
's internal graphics display and wasn't of much utility outside of that context.
-
Instead of cluttering the terminal with messages,
pairwise_comparisons
function now instead adds two columns (test.details
andp.value.adjustment
) to all outputs specifying which test was carried out and which adjustment method is being used for p-value correction. -
Gets rid of
groupedstats
andcrayon
from dependencies.
- With
jmv 1.0.8
, the results from the Dwass-Steel-Crichtlow-Fligner test will be slightly different.
-
The
p.value.label
in the output dataframe has been renamed tolabel
to consider the possibility that Bayes Factor tests might also be supported in future. -
The label now specified whether the p-value was adjusted or not for multiple comparisons.
- First release of the package.