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.
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.
stat_smooth()to reduce messages about this.
corplot()uses a more color blind friendly palette and defaults to showing correlations and p-values.
cowplotfor themes and arranging multiple plots.
modelTest()now works correctly when there are interaction terms with categorical variables and when most “on-the-fly” transformations are performed, such as
log()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.
data.tableto the DESCRIPTION and dependencies.
residualDiagnostics.lm()and so too
modelDiagnostics.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.
JWileymischave been separated into other packages, including the new
extraoperatorspackage, 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 using
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.
detailedTests()is now more generic and dispatches to
.detailedTestsVGLM()to provide detailed tests for both linear mixed effects models and multinomial logistic regression models fit by
ezMULTINOM()is now deprecated in favor of the new, more generic
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 by
compareLMER() A function to compare two models estimated by
lmer() include significance tests and effect sizes for estimates of the variance explained.
detailedTests() A function to compute detailed tests on a model estimated from
lmer() 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 from
detailedTests() based on
lmer() 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).
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 old
mvqq() function is now deprecated.
testdistr() includes optional robust estimates for univariate and multivariate normal data
logicals A 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,
testdistr() function to plot data against different theoretical distributions using
ggplot2. A sort of generalized
qqnorm() allowing other distributions besides the normal distribution.
winsorizor() Function moved from
pscore package. Sets any values beyond specific quantiles of the empirical data to the specified quantiles. Can work on vectors, data frames, or matrices.