corplot()and functions that build on it (likeplot()methods forSEMSummary()now use a 45 degree angle for column labels to improve readability with long variable names. As always this is aggplot2theme option and can be overwritten by the user.
geom_tufterange()simple geom to add a left and bottom line that shows the observed range of non missing data.theme_tufte()simple mostly empty theme built off the minimal theme.
- Updated code in preparation for is.atomic(NULL) no longer working.
- Updated to address aes_string() being deprecated
is.naz()is a new function that is the testing counter part tonaz.omit(). Notably, bothis.naz()andnaz.omit()both also identify and exclude non finite values now. This is new behavior.
egltable()has more error checks including for variables that are all missing values. Frequency (percent) results are now shown more clearly with a percent sign (%). This corresponds to changes in the backend ofegltable()to facilitate more tests and descriptives to be calculated.
diffCircular()calculates the circular difference between two vectors.saveRDSfst()saves RDS files usingfstfor multithreaded compression.readRDSfst()reads RDS files usingfstfor multithreaded decompression.
egltable()now correctly handles categorical variables by a grouping variable, when the categorical variable is not a factor class. Fixes a bug that could occur with cells with zero frequencies when the variables were not factors.winsorizor()would not properly check if a vector of percentiles was passed. This is fixed and a new test added to prevent regression.
- Now using testthat 3e and preferably for light/dark package website.
meanCircular()would be off by pi in some circumstances. This has been corrected and more test cases added.SEMSummary()pairwise correlations were based on the standardized pairwise covariance matrix, which used the same standard deviation for a variable regardless of the pair. This has now been fixed.- Formula has been specified in calls to
stat_smooth()to reduce messages about this.
corplot()uses a more color blind friendly palette and defaults to showing correlations and p-values.- Model diagnostics are smarter about linear regressions with essentially discrete model predictions and no longer try to run quantile regression to examine homogeneity of variance in these cases.
- Switched to using
ggpubrinstead ofcowplotfor themes and arranging multiple plots. - Bumped minimum version of
Rrequired. - Further unit testing.
modelTest()now works correctly when there are interaction terms with categorical variables and when most "on-the-fly" transformations are performed, such aslog()etc. Does not work when new variables that are a composite of multiple variables are created, e.g.,I(hp + wt), but a more informative error message is given.- Added the correct version requirements for
data.tableto the DESCRIPTION and dependencies.
- Polished methods for functions, including
APAStyler()methods.
residualDiagnostics.lm()and so toomodelDiagnostics.lmwould return the index of extreme values based on the complete data used for modelling, not of the original input dataset. This made it difficult to identify and remove extreme values in subsequent model runs. This is now corrected.
- This version includes a significant re-write of the package that
results in many of the previous user-facing functions changing
names and arguments. This will likely break old code.
This re-write was necessary to help standardize functions and
arguments, and to make functions more robust.
Many functions are now generics with specific methods written.
Further, some functions previously bundled into
JWileymischave been separated into other packages, including the newextraoperatorspackage, covering binary operators, and a package for diagnostics on mixed models. See the new vignettes added to the package for examples of current practice in usingJWileymisc.
egltable()has added statistical tests for paired data. For continuous, parametric paired data, a pseudo Cohen's d is calculated on the change scores.
-
omegaSEM()Function that calculates coefficient omega for measuring internal consistency reliability. Works for two level models and returns within and between level omega values. -
egltable()Function has added effect sizes when multiple groups are compared including Cohen's d for two groups, eta-squared for multiple groups, and phi for categorical variables.
testdistr()now only finds extreme values for the right tail of a chi-square distribution..detailedTestsVGLM()now identifies levels of the outcome correctly.
detailedTests()is now more generic and dispatches to.detailedTestsLMER()or.detailedTestsVGLM()to provide detailed tests for both linear mixed effects models and multinomial logistic regression models fit byvglm().ezMULTINOM()is now deprecated in favor of the new, more genericdetailedTests().
-
testdistr()now creates more appropriate plots for discrete distributions including the Poisson and Negative Binomial. -
moments()now updated to accomodate changes in the lavaan package (thanks to Yves Rosseel) -
TukeyHSDgg()updated to use the emmeans package instead of the now defunct lsmeans package. -
formatLMER()returned the lower confidence interval twice instead of the lower and upper confidence interval. This is now fixed.
-
R2LMER()A simple function to calculate the marginal and conditional variance accounted for by a model estimated bylmer(). -
compareLMER()A function to compare two models estimated bylmer()include significance tests and effect sizes for estimates of the variance explained. -
detailedTests()A function to compute detailed tests on a model estimated fromlmer()including confidence intervals for parameters, significance tests, where possible, overall model fit, and effect sizes for the model and each variable. -
formatLMER()A function to nicely format detailed model results, possibly from multiple models. Requires results fromdetailedTests()based onlmer()models, at the moment. -
iccMixed()A function to calculate the intraclass correlation coefficient using mixed effects models. Works with either normally distributed outcomes or binary outcomes, in which case the latent variable estimate of the ICC is computed. -
nEffective()Calculates the effective sample size based on the number of independent units, number of observations per unit, and the intraclass correlation coefficient. -
acfByID()Calculates the lagged autocorrelation of a variable by an ID variable and returns a data.table for further use, such as examination, summary, or plotting -
meanDeviations()A simple function to calculate means and mean deviations, useful for creating between and within versions of a variable in a data.table -
as.na()function added to convert data to missing (NA) while preserving the class/type of the data (useful for data.table). -
meanDecompose()function added to decompose multilevel or repeated measures data into means and residuals. -
timeshift()function added to center a time variable at a new zero point. Useful when times may start and end off the standard 24 hour period (e.g., 11am to 2am, which technically fall on different dates). -
intSigRegGraph()function added to graph regions of significance from interactions with linear models as well as the mostly helper function,findSigRegions(). -
ezMULTINOM()new function added to make running multinomial logistic regression easy in R, along with all pairwise contrasts and omnibus tests of statistical significance. -
testdistr()function expanded to cover multivariate normal data, and the oldmvqq()function is now deprecated. -
testdistr()includes optional robust estimates for univariate and multivariate normal data -
formatHtest()gains support for pearson, kendal, and spearman correlations from thecor.test()function -
logicalsA series of support functions for findings values in a particular range, such as%gele%for values greater than or equal to the min and less than or equal to the max as well as to automatically subset the data when prefixed with an s,%sgele%%sin%etc.
-
winsorizor()now properly handles atomic data. Fixes an issue where variables in a data table would be atomic after calling thescale()function andwinsorizor()would fail. -
egltable()now works with data.tables
-
testdistr()function to plot data against different theoretical distributions usingggplot2. A sort of generalizedqqnorm()allowing other distributions besides the normal distribution. -
winsorizor()Function moved frompscorepackage. Sets any values beyond specific quantiles of the empirical data to the specified quantiles. Can work on vectors, data frames, or matrices.
Initial release to CRAN.