Please check the latest news (change log) and keep this package updated.
import() support for importing URL-source data
files and no-extension files.paired.d.type argument for TTEST():
Allow for specifying 3 types of Cohen’s d for paired-samples
t-test ("dz", "dav", and
"drm"). See Lakens (2013) for details.
"dz" (d for standardized difference)
\[\text{Cohen's } d_{z} = \frac{M_{diff}}{SD_{diff}}\]
"dav" (d for average standard
deviation)
\[\text{Cohen's } d_{av} = \frac{M_{diff}}{\frac{SD_{1} + SD_{2}}{2}}\]
"drm" (d for repeated measures, corrected
for correlation)
\[\text{Cohen's } d_{rm} = \frac{M_{diff} \times \sqrt{2(1 - r_{1,2})}}{\sqrt{SD_{1}^2 + SD_{2}^2 - 2 \times r_{1,2} \times SD_{1} \times SD_{2}}}\]
library(bruceR).add() and added(): Enhanced
functions designed to create, modify, and/or delete variables. The
functions combine the advantages of :=
(data.table), mutate() (dplyr), and
transmute() (dplyr). See help
page for the usage and convenience..sum() and .mean(): Tidy
version of SUM() and MEAN() designed only for
add() and added(). See help
page for the usage and convenience.SUM(),
MEAN(), EMMEANS().pkgdown. Configuration is specified in
_pkgdown.yml.R CMD check workflow on GitHub, which checks
the code for each push.cc(). Now it becomes much more
convenient!MANOVA(): “Data are aggregated to mean (across
items/trials) if there are >=2 observations per subject and cell. You
may use Linear Mixed Model to analyze the data, e.g., with subjects and
items as level-2 clusters.”center argument of PROCESS()
(default is TRUE) for users who want to turn off the
automatic grand-mean centering. However, mean centering is still highly
suggested if one aim to obtain “main effect” rather than “fixed effect”
(note: a fixed effect is not necessarily a main effect).estimator argument of CFA()
(default is "ML") for users who want to use any other
estimator (fixed issue #17).cc(): Split up a string (with separators)
into a character vector (whitespace around separator is trimmed). For
example, cc("A 1 , B 2 ; C 3 | D 4 \t E 5") produces a
vector of c("A 1", "B 2", "C 3", "D 4", "E 5"). The default
separators include , ; | \n \t. Users may also specify a
separator.Added a guideline and examples for creating interaction plots
using the returned object of MANOVA() and
EMMEANS(). You can save the returned object and use the
emmeans::emmip() function to create an interaction plot
(based on the fitted model and a formula specification). For usage,
please see the help page of emmeans::emmip(). It returns an
object of class ggplot, which can be easily modified and
saved using ggplot2 syntax.
Added an explanation of the automatic grand-mean centering in
PROCESS().
afex package would see
an unusual error when using the EMMEANS() function
(Error: $ operator is invalid for atomic vectors). So now
afex is again a strong dependency of
bruceR, such that it is automatically installed when
installing bruceR.EMMEANS() (when
model is null).PROCESS() when
setting mod.type="3-way" for multilevel models.print_table().PROCESS()
for models without "x-y" in mod.path (e.g.,
Model 7).set.wd() in an R Markdown
file.bruceR when
library(bruceR).phia::testInteractions() in the
output of EMMEANS(). These tests produce identical results
to those obtained with the SPSS GLM (/EMMEANS) syntax.Freq(): Now both vector and
data frame can be used. For example, users may specify either
Freq(data$variable) or
Freq(data, "variable").GLM_summary() and
HLM_summary().HLM_summary():
level2.predictors and vartypes.lmerTest, mediation,
interactions, and lavaan are now strong
dependencies such that they would be installed when installing
bruceR. This also fixes an error when using
PROCESS() without these packages installed.print_table().GLM_summary() and
HLM_summary() when only one factor-type predictor with
>= 3 levels is in a regression model. Other bugs in these two
functions have also been fixed.PROCESS() when
setting mod.path="all" in testing multilevel moderated
mediation effects. Fixed another bug of CI output for direct effects
when testing multilevel models.TTEST(): One-sample, independent-samples,
and paired-samples t-test. Multiple dependent/independent
variables can be tested simultaneously. It also tests the assumption of
homogeneity of variance and allows users to determine whether variances
are equal or not. Cohen’s d and 95% CI are reported by default
(see Details and Examples in its help page for an issue about the
inconsistency in the results of 95% CI of Cohen’s d
between R packages). Bayes factor BF10 is also supported. Key
results can be saved in APA format to MS Word.import() / export():
Import/export data from/to a file with the two tidy functions, relieving
users of the burden of remembering lots of read_xxx() /
write_xxx() functions. Many file formats are supported
(especially .txt, .csv, .tsv, .psv, .xls, .xlsx, .sav, .dta, .rda,
.rdata, and clipboard). Note that the two functions are inspired by
rio::import() / rio::export() and have several
modifications for more convenient use. Since this version, the package
rio is no longer a strong dependency of bruceR
and would not be loaded when loading bruceR.Improved MANOVA() and EMMEANS():
print_table().EMMEANS() adopts this reasonable approach. If one uses
MANOVA() and EMMEANS() to conduct the same
t-test as using the TTEST() function, the results
will be identical. Indeed, the estimation methods of Cohen’s d
in t-tests are acknowledged. In computing pooled SD in
ANOVAs, it uses (1) the square root of mean square
error (MSE) for between-subjects designs and (2)
the square root of mean variance of all paired differences of the
residuals of repeated measures for within-subjects and mixed
designs. In both situations, it extracts the lm object from
the returned value of MANOVA(). Then, it mainly uses the
sigma() and residuals() functions,
respectively, to do these estimates. For source code, see R file on
GitHub. Thus, the results of Cohen’s d for designs with
repeated measures are now different from those in bruceR
old versions (< 0.8.0), which indeed used an inappropriate method to
compute pooled SD in such designs.ss.type for MANOVA()
to specify either Type-II or Type-III Sum of Square; (2)
aov.include for MANOVA() and
model.type for EMMEANS(), for details, see the
help pages.Improved Alpha(): Now it directly uses
psych::alpha() and psych::omega(), rather than
jmv::reliability(), to perform reliability analysis. The
format of result output has been changed and improved.
Improved EFA() (almost completely rewritten): Now it
directly uses psych::principal() and
psych::fa(), rather than jmv::efa(), to
perform factor analysis (PCA or EFA). The format of result output has
been changed and improved. MS Word output has been supported. A wrapper
function PCA() has been added:
EFA(..., method="pca").
Improved CFA() and lavaan_summary():
Now CFA() only uses the lavaan::cfa(), rather
than jmv:cfa(), to build model, and then uses
lavaan_summary() to present results. For
lavaan_summary(), many bugs have been fixed, and the format
of result table has been changed and improved. Both functions now
support saving table to MS Word.
Package dependencies: Much fewer strong dependencies, for faster
and more robust installation. Removed rio and
jmv from dependencies. No longer load rio and
psych when library(bruceR).
set_wd() for set.wd().print_table(): Fixed an issue of incorrect
length of Chinese character output in print_table().
Between-column blanks are now 2 spaces (rather than 1 space) for a
clearer presentation of table columns.lavaan_summary() and
granger_test().digits parameter as the equivalent to the
nsmall parameter for all relevant functions.mediation, interactions,
MuMIn, and texreg are now SUGGESTS rather than
IMPORTS.PROCESS().Corr().PROCESS(): PROCESS for mediation,
moderation, and conditional process (moderated mediation) analyses! This
function supports a total of 24 kinds of SPSS PROCESS models (Hayes,
2018) and also supports multilevel mediation/moderation analyses.
Overall, it supports the most frequently used types of mediation,
moderation, moderated moderation (3-way interaction), and moderated
mediation (conditional indirect effect) analyses for (generalized)
linear or linear mixed models. Regression model summary and effect
estimates (simple slopes and/or indirect effects) are printed in an
elegant way.lavaan_summary(): Tidy report of lavaan
model.RANDBETWEEN() function.print_table() and other
functions using print_table() inside:
Describe(), Freq(), Corr(),
MANOVA(), med_summary(),
granger_causality().EMMEANS(): There is considerable
disagreement on how to compute Cohen’s d. Users should not take
the default output as the only right results and are completely
responsible for setting the “sd.pooled”.model_summary(): (1) Model names with
NULL; (2) Multicollinearity check results with
NULL or other problems; (3) UTF-8 encoding problem in WPS
software (no such problem in Microsoft Word).granger_causality(): Granger causality
test (multivariate) based on vector autoregression (VAR) model. This
function is an advanced and more general version of the function
granger_test() (bivariate).set.wd(): Now it uses
rstudioapi::getSourceEditorContext() to extract file path
(even effective when running in R console), which only requires RStudio
version >= 0.99.1111 and no longer has encoding problems (see release
note in 0.6.1).theme_bruce(): Now it uses
ggtext::element_markdown() to render Markdown/HTML rich
text format, which can be used in plot text (e.g., titles).EMMEANS(): Now its results are always
identical to those in SPSS (by setting model="multivariate"
in emmeans::joint_tests() and
emmeans::emmeans(), which use the lm or
mlm objects rather than the aov object to
perform tests). For a few cases with singular error matrix (i.e., some
variables are linearly dependent), the results of simple-effect
F tests will not be reported, but estimated marginal means and
pairwise comparisons are not affected and so are still reported. Note
that the EMMEANS results in old versions of
bruceR (version < 0.6.0) were identical to SPSS, but
version 0.6.0 deprecated the parameter repair and no longer
set model$aov=NULL, which made the results not identical to
SPSS (particularly for ANOVAs with repeated measures). In response to a
user’s feedback, now 0.6.2 has improved this function and makes its
results accurate again.\code{\link[package:function]{package::function()}}.CFA() (for lavaan-style output).HLM_ICC_rWG(): Tidy report of HLM indices
“ICC(1)” (non-independence of data), “ICC(2)” (reliability of group
means), and “rWG”/“rWG(J)” (within-group agreement for
single-item/multi-item measures).Run(): Run code parsed from text.show_colors(): Show multiple colors (or a
palette) in a plot.%^%: Paste strings together (a wrapper of
paste0()).set.wd(): Now it converts the extracted path
string from “UTF-8” to “GBK” on Windows system to support paths
including Chinese characters (otherwise, the path would become messy
code and cause an error). Note that this problem does not exist on Mac
OS. In addition, warning messages will be printed into the console if
the user’s RStudio version is lower than required (RStudio version >=
1.4.843 is required for a complete implementation of this
function).Alpha(): Now it adds a parameter
varrange (to keep the same as SUM(),
MEAN(), …) and reports both Cronbach’s α and McDonald’s ω,
with more detailed documentation.Three ways to specify the variable list (implemented in the functions such as
SUM(),MEAN(),Alpha()):1.
var + items: use the common and unique parts of variable names. (e.g.,var="RSES", items=1:10, rev=c(3, 5, 8, 9, 10))2.
vars: directly define the variable list. (e.g.,vars=c("E1", "E2", "E3", "E4", "E5"), rev=c("E1", "E2"))3.
varrange: use the start and end positions of the variable list. (e.g.,varrange="E1:E5", rev=c("E1", "E2"))
Corr() (relevant to the
changes in psych::corr.test() in a forthcoming release of
the psych package).0 errors √ | 0 warnings √ | 0 notes √model_summary(): Tidy report of
(single/multiple) regression models (into console or to a Word/HTML
file; supporting most types of models; based on the texreg
package).med_summary(): Tidy report of
(simple/moderated) mediation analyses (based on the
mediation package).ccf_plot: Cross-correlation analysis
(plotting with ggplot2).granger_test: Granger test of predictive
causality (based on the lmtest::grangertest()
function).set.wd(),
Describe(), Corr(), MANOVA(), and
EMMEANS().library(bruceR).pkg_install_suggested()
function).bruceR:
BRoadly Useful
Convenient and Efficient
R functions that BRing
Users Concise and
Elegant R data analyses.MANOVA() and EMMEANS():
ANOVA, simple-effect analyses, and multiple comparisons (based on the
afex and emmeans packages).