tidyBF: Tidy Wrapper for BayesFactor Package

Package Status Usage GitHub References
CRAN_Release_Badge Travis Build Status Daily downloads badge GitHub version Website
CRAN Checks AppVeyor Build Status Weekly downloads badge Forks Features
minimal R version lifecycle Monthly downloads badge Github Issues vignettes
GitHub code size in bytes Coverage Status Total downloads badge Github Stars DOI
Licence Codecov test coverage HitCount Last-changedate GitHub last commit
status R build status Gitter chat Project Status contributions welcome

Overview

tidyBF package is a tidy wrapper around the BayesFactor package that always expects the data to be in the tidy format and return a tibble containing Bayes Factor values. Additionally, it provides a more consistent syntax and by default returns a dataframe with rich details. These functions can also return expressions containing results from Bayes Factor tests that can then be displayed in custom plots.

Installation

To get the latest, stable CRAN release:

install.packages("tidyBF")

You can get the development version of the package from GitHub. To see what new changes (and bug fixes) have been made to the package since the last release on CRAN, you can check the detailed log of changes here: https://indrajeetpatil.github.io/tidyBF/news/index.html

If you are in hurry and want to reduce the time of installation, prefer-

# needed package to download from GitHub repo
install.packages("remotes")

remotes::install_github(
  repo = "IndrajeetPatil/tidyBF", # package path on GitHub
  quick = TRUE # skips docs, demos, and vignettes
)

If time is not a constraint-

remotes::install_github(
  repo = "IndrajeetPatil/tidyBF", # package path on GitHub
  dependencies = TRUE, # installs packages which `tidyBF` depends on
  upgrade_dependencies = TRUE # updates any out of date dependencies
)

Citation

This package is one component of the ggstatsplot package. If you want to cite this package in a scientific journal or in any other context, run the following code in your R console:

citation("tidyBF")
#> 
#>   Patil, I. (2018). ggstatsplot: 'ggplot2' Based Plots with Statistical
#>   Details. CRAN. Retrieved from
#>   https://cran.r-project.org/web/packages/ggstatsplot/index.html
#> 
#> A BibTeX entry for LaTeX users is
#> 
#>   @Article{,
#>     title = {{ggstatsplot}: 'ggplot2' Based Plots with Statistical Details},
#>     author = {Indrajeet Patil},
#>     year = {2018},
#>     journal = {CRAN},
#>     url = {https://CRAN.R-project.org/package=ggstatsplot},
#>     doi = {10.5281/zenodo.2074621},
#>   }

Summary of available tests

Behind the curtains, tidyBF provides an easier syntax to marry functionalities provided by the following two packages in a unified framework:

Analysis Function Hypothesis testing Estimation Function
(one/two-sample) t-test bf_ttest Yes Yes BayesFactor::ttestBF + bayestestR::describe_posterior
one-way ANOVA bf_oneway_anova Yes Yes BayesFactor::anovaBF + performance::r2_bayes
correlation bf_corr_test Yes Yes BayesFactor::correlationBF + bayestestR::describe_posterior
(one/two-way) contingency table bf_contingency_tab Yes Yes BayesFactor::contingencyTableBF + effectsize::effectsize
random-effects meta-analysis bf_meta_random Yes Yes metaBMA::meta_random

Notation

The results are always displayed as a Bayes Factor in favor of the null hypothesis over the alternative hypothesis. Additionally, the values are logged to avoid huge numbers. Therefore, the notation is: log_{e}(BF_{01}).

Also, please note that this makes flipping the evidence easy: log_{e}(BF_{10}) = - log_{e}(BF_{01})

Benefits

Below are few concrete examples of where tidyBF wrapper might provide a more friendly way to access output from or write functions around BayesFactor.

Syntax consistency

BayesFactor is inconsistent with its formula interface. tidyBF avoids this as it doesn’t provide the formula interface for any of the functions.

# setup
set.seed(123)

# with `BayesFactor` ----------------------------------------
suppressPackageStartupMessages(library(BayesFactor))
data(sleep)

# independent t-test: accepts formula interface
ttestBF(formula = wt ~ am, data = mtcars)
#> Bayes factor analysis
#> --------------
#> [1] Alt., r=0.707 : 1383.367 ±0%
#> 
#> Against denominator:
#>   Null, mu1-mu2 = 0 
#> ---
#> Bayes factor type: BFindepSample, JZS

# paired t-test: doesn't accept formula interface
ttestBF(formula = extra ~ group, data = sleep, paired = TRUE)
#> Error in ttestBF(formula = extra ~ group, data = sleep, paired = TRUE): Cannot use 'paired' with formula.

# with `tidyBF` ----------------------------------------
library(tidyBF)

# independent t-test
bf_ttest(data = mtcars, x = am, y = wt)
#> # A tibble: 1 x 13
#>   term       estimate conf.low conf.high    pd rope.percentage
#>   <chr>         <dbl>    <dbl>     <dbl> <dbl>           <dbl>
#> 1 Difference    -1.26    -1.79    -0.722     1               0
#>   prior.distribution prior.location prior.scale effect component    bf10
#>   <chr>                       <dbl>       <dbl> <chr>  <chr>       <dbl>
#> 1 cauchy                          0       0.707 fixed  conditional 1383.
#>   log_e_bf10
#>        <dbl>
#> 1       7.23

# paired t-test
bf_ttest(data = sleep, x = group, y = extra, paired = TRUE, subject.id = ID)
#> # A tibble: 1 x 13
#>   term       estimate conf.low conf.high    pd rope.percentage
#>   <chr>         <dbl>    <dbl>     <dbl> <dbl>           <dbl>
#> 1 Difference     1.40    0.508      2.31 0.996               0
#>   prior.distribution prior.location prior.scale effect component    bf10
#>   <chr>                       <dbl>       <dbl> <chr>  <chr>       <dbl>
#> 1 cauchy                          0       0.707 fixed  conditional  17.3
#>   log_e_bf10
#>        <dbl>
#> 1       2.85

Expressions for plots

Although all functions default to returning a dataframe, you can also use it to extract expressions that can be displayed in plots.

t-test

# setup
set.seed(123)
library(ggplot2)

# using the expression to display details in a plot
ggplot(ToothGrowth, aes(supp, len)) +
  geom_boxplot() + # two-sample t-test results in an expression
  labs(subtitle = bf_ttest(ToothGrowth, supp, len, output = "expression"))

anova

# setup
set.seed(123)
library(ggplot2)
library(ggforce)
library(tidyBF)

# plot with subtitle
ggplot(iris, aes(x = Species, y = Sepal.Length)) +
  geom_violin() +
  geom_sina() +
  labs(subtitle = bf_oneway_anova(iris, Species, Sepal.Length, output = "expression"))

correlation test

# setup
set.seed(123)
library(ggplot2)
library(tidyBF)

# using the expression to display details in a plot
ggplot(mtcars, aes(wt, mpg)) +
  geom_point() +
  geom_smooth(method = "lm") +
  labs(subtitle = bf_corr_test(mtcars, wt, mpg, output = "expression"))
#> `geom_smooth()` using formula 'y ~ x'

contingency tabs analysis

# setup
set.seed(123)
library(ggplot2)
library(tidyBF)

# basic pie chart
ggplot(as.data.frame(table(mpg$class)), aes(x = "", y = Freq, fill = factor(Var1))) +
  geom_bar(width = 1, stat = "identity") +
  theme(axis.line = element_blank()) +
  # cleaning up the chart and adding results from one-sample proportion test
  coord_polar(theta = "y", start = 0) +
  labs(
    fill = "Class",
    x = NULL,
    y = NULL,
    title = "Pie Chart of class (type of car)",
    subtitle = bf_contingency_tab(as.data.frame(table(mpg$class)), Var1, counts = Freq, output = "h1")
  )

meta-analysis

# setup
set.seed(123)
library(metaviz)
library(ggplot2)

# meta-analysis forest plot with results random-effects meta-analysis
viz_forest(
  x = mozart[, c("d", "se")],
  study_labels = mozart[, "study_name"],
  xlab = "Cohen's d",
  variant = "thick",
  type = "cumulative"
) + 
  labs(
    title = "Meta-analysis of Pietschnig, Voracek, and Formann (2010) on the Mozart effect",
    subtitle = bf_meta_random(
      data = dplyr::rename(mozart, estimate = d, std.error = se),
      output = "expression",
      metaBMA.args = list(rscale_discrete = 0.880),
      conf.level = 0.99
    )
  ) +
  theme(text = element_text(size = 12))

Convenient way to extract detailed output from BayesFactor objects

The package provides bf_extractor function to conveniently extract important details from these objects:

# setup
set.seed(123)
library(tidyBF)
library(BayesFactor)
data(puzzles)

# model
result <-
  anovaBF(
    RT ~ shape * color + ID,
    data = puzzles,
    whichRandom = "ID",
    whichModels = "top",
    progress = FALSE
  )

# extract details
bf_extractor(result)
#> # A tibble: 21 x 20
#>    term                estimate conf.low conf.high    pd rope.percentage effect
#>    <chr>                  <dbl>    <dbl>     <dbl> <dbl>           <dbl> <chr> 
#>  1 mu                    45.0    43.7      46.4    1               0     fixed 
#>  2 shape-round            0.429   0.0643    0.801  0.992           0.141 fixed 
#>  3 shape-square          -0.429  -0.801    -0.0643 0.992           0.141 fixed 
#>  4 color-color           -0.426  -0.799    -0.0461 0.990           0.162 fixed 
#>  5 color-monochromatic    0.426   0.0461    0.799  0.990           0.162 fixed 
#>  6 ID-1                   2.47    0.783     4.37   0.995           0     random
#>  7 ID-2                   0.439  -1.21      2.20   0.698           0.231 random
#>  8 ID-3                   0.907  -0.849     2.66   0.848           0.156 random
#>  9 ID-4                   0.466  -1.47      2.20   0.704           0.218 random
#> 10 ID-5                   3.17    1.38      5.00   0.999           0     random
#>    component    bf10 log_e_bf10    r2 std.dev ci.width r2.conf.low r2.conf.high
#>    <chr>       <dbl>      <dbl> <dbl>   <dbl>    <dbl>       <dbl>        <dbl>
#>  1 extra       2.65       0.974 0.732  0.0509       95       0.609        0.812
#>  2 conditional 0.233     -1.45  0.732  0.0509       95       0.609        0.812
#>  3 conditional 0.239     -1.43  0.732  0.0509       95       0.609        0.812
#>  4 conditional 2.65       0.974 0.732  0.0509       95       0.609        0.812
#>  5 conditional 0.233     -1.45  0.732  0.0509       95       0.609        0.812
#>  6 conditional 0.239     -1.43  0.732  0.0509       95       0.609        0.812
#>  7 conditional 2.65       0.974 0.732  0.0509       95       0.609        0.812
#>  8 conditional 0.233     -1.45  0.732  0.0509       95       0.609        0.812
#>  9 conditional 0.239     -1.43  0.732  0.0509       95       0.609        0.812
#> 10 conditional 2.65       0.974 0.732  0.0509       95       0.609        0.812
#>    r2.component prior.parameter prior.distribution prior.location prior.scale
#>    <chr>        <chr>           <chr>                       <dbl>       <dbl>
#>  1 conditional  fixed           cauchy                          0         0.5
#>  2 conditional  fixed           cauchy                          0         0.5
#>  3 conditional  fixed           cauchy                          0         0.5
#>  4 conditional  fixed           cauchy                          0         0.5
#>  5 conditional  fixed           cauchy                          0         0.5
#>  6 conditional  fixed           cauchy                          0         0.5
#>  7 conditional  fixed           cauchy                          0         0.5
#>  8 conditional  fixed           cauchy                          0         0.5
#>  9 conditional  fixed           cauchy                          0         0.5
#> 10 conditional  fixed           cauchy                          0         0.5
#> # ... with 11 more rows

Acknowledgments

The hexsticker was generously designed by Sarah Otterstetter (Max Planck Institute for Human Development, Berlin).

Code of Conduct

Please note that the tidyBF project is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.