ggcommonality creates bar plots of unique and joint effects from a
commonality analysis of a linear regression model. The S4 branch extends
its functionality using an object-oriented syntax. The package calls on
yhat to perform commonality analyses (Nimon, Oswald, and Roberts.
2023), building bar plots in the style of those appearing in the MAPLE
Lab’s work applying commonality analysis to the
compositions of Bach and Chopin (Delle Grazie, Anderson, and Schutz
2025; Anderson and Schutz 2022).
Partitions are plotted sequentially in alphabetical order, starting with unique effects and are built iteratively with joint effects at higher orders on top.
Note: This is my first R package and is currently under active development, so there are likely several bugs that remain to be found and fixed.
You can install ggcommonality (under development) from GitHub with:
devtools::install_github("cmndrsn/ggcommonality", build_vignettes = TRUE, force = TRUE)The function produces a GGCommonality object, containing ingredients for plotting commonality effects, along with the results from the commonality analysis.
library(ggcommonality)
my_formula <- mpg ~ wt + hp
p <- ggcom(formula = my_formula,
data = mtcars)
#> Registered S3 method overwritten by 'mosaic':
#> method from
#> fortify.SpatialPolygonsDataFrame ggplot2
p |> attributes() |> summary()
#> Length Class Mode
#> data 11 data.frame list
#> data.boot 400 -none- numeric
#> formula 3 formula call
#> stack 1 -none- name
#> n_replications 1 -none- numeric
#> sample_column 1 -none- name
#> resample_type 1 -none- character
#> wild_type 1 -none- character
#> include_total 1 -none- logical
#> ci_bounds 2 -none- numeric
#> seed 1 -none- name
#> ... 1 -none- name
#> class 1 -none- characterggcommonality objects can be defined, plotted, and summarized using built-in methods.
# visualize commonality effects and add confidence intervals
plot(p) +
ggcom_ci(p)
#> Warning: Using `size` aesthetic for lines was deprecated in ggplot2 3.4.0.
#> ℹ Please use `linewidth` instead.
#> ℹ The deprecated feature was likely used in the ggcommonality package.
#> Please report the issue to the authors.
#> This warning is displayed once per session.
#> Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
#> generated.
#> Bootstrapped confidence intervals:
#> # A tibble: 3 × 3
#> # Groups: com [3]
#> com lci uci
#> <fct> <dbl> <dbl>
#> 1 hp 0.0291 0.169
#> 2 wt 0.0830 0.346
#> 3 wt,hp 0.427 0.691
We can also check the output from the yhat package
ggcom_yhat(p)
#> $yhat
#> $yhat$LM_Output
#>
#> Call:
#> lm(formula = x@formula, data = x@data)
#>
#> Residuals:
#> Min 1Q Median 3Q Max
#> -3.941 -1.600 -0.182 1.050 5.854
#>
#> Coefficients:
#> Estimate Std. Error t value Pr(>|t|)
#> (Intercept) 37.22727 1.59879 23.285 < 2e-16 ***
#> wt -3.87783 0.63273 -6.129 1.12e-06 ***
#> hp -0.03177 0.00903 -3.519 0.00145 **
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#>
#> Residual standard error: 2.593 on 29 degrees of freedom
#> Multiple R-squared: 0.8268, Adjusted R-squared: 0.8148
#> F-statistic: 69.21 on 2 and 29 DF, p-value: 9.109e-12
#>
#>
#> $yhat$Beta_Weights
#> wt hp
#> -0.6295545 -0.3614507
#>
#> $yhat$Structure_Coefficients
#> wt hp
#> [1,] -0.9542295 -0.8536101
#>
#> $yhat$Commonality_Data
#> $yhat$Commonality_Data$CC
#> Coefficient % Total
#> Unique to wt 0.2243 27.13
#> Unique to hp 0.0740 8.94
#> Common to wt, and hp 0.5285 63.92
#> Total 0.8268 100.00
#>
#> $yhat$Commonality_Data$CCTotalbyVar
#> Unique Common Total
#> wt 0.2243 0.5285 0.7528
#> hp 0.0740 0.5284 0.6024
#>
#>
#> $yhat$Effect_Size
#> Effect.Size Recommended
#> Wherry1 0.808226750298805 No
#> Claudy3 0.824785233237495 No
#> Smith 0.808866705525839 No
#> Wherry2 0.814839620978156 No
#> Olkin & Pratt 0.818527562995977 Yes
#> Pratt 0.818271515745636 No
#>
#> $yhat$Comment
#> [1] "The Effect Size recommendations are based on Yin and Fan (2001). Your dataset may take on a different covariance structure, thus making another effect size estimate more appropriate."
#>
#>
#> $ci
#> Unique to wt Unique to hp Common to wt, and hp
#> 2.5% 0.0830375 0.029070 0.4274225
#> 97.5% 0.3456300 0.169165 0.6907550
#> Total
#> 2.5% 0.7420675
#> 97.5% 0.9258700Commonality effects can be stacked in multiple ways:
# update object
p@stack <- 'common'
# stack
plot(p) +
ggcom_ci(p)+
ylim(0, 0.7)
#> Bootstrapped confidence intervals:
#> type lower upper
#> 2.5% unique 0.1578475 0.4153675
#> 2.5%1 common 0.4274225 0.6907550
# update object
slot(p, 'stack') <- "partition"
# stack
plot(p) +
ggcom_ci(p)
#> Bootstrapped confidence intervals:
#> category lower upper
#> 2.5% wt 0.6116075 0.8537450
#> 2.5%1 hp 0.5013850 0.7596975
The ggcom_ci() method prints confidence intervals generated for
stacked effects.
ggcom_ci(p)
#> Bootstrapped confidence intervals:
#> category lower upper
#> 2.5% wt 0.6116075 0.8537450
#> 2.5%1 hp 0.5013850 0.7596975
#> mapping: x = ~x_mid, ymin = ~lower, ymax = ~upper
#> geom_errorbar: na.rm = FALSE, orientation = NA, lineend = butt, width = 0.3
#> stat_identity: na.rm = FALSE
#> position_identityThis package uses percentile-based confidence intervals.
The resample_type argument specifies whether to generate random-x,
confidence intervals, fixed-x, or wild-x confidence intervals. The
appendices to Fox and Weisberg
(2018) summarizes the advantages and disadvantages of fixed
vs. random-x bootstrapping. There is also a wild-x option
implemented as
describedhere.
If stack_by = "partition", confidence intervals sum the
unique and joint effects for individual commonality partitions achieved through bootstrapping.
Otherwise, if stack_by = "common", separate confidence intervals are
generated for the sum of unique effects and the sum of joint effects.
Confidence intervals by default for these options are based on positive
coefficients, but can be specified with an argument to ggcom_ci()
(sign = “+” for positive coefficients, sign = “-” for negative ones, or
sign = “” for both).
# set r's random number generator
p1 <- ggcom(
formula = my_formula,
data = mtcars,
resample_type = "fixed"
)
p2 <- ggcom(
formula = my_formula,
data = mtcars,
resample_type = "wild"
)
library(patchwork)
# stack
plot(p1) +
ggcom_ci(p1)+
ylim(0,0.65)+
ggtitle("Fixed")|
plot(p2) +
ggcom_ci(p2)+
ylim(0,0.65)+
ggtitle("Wild")
#> # A tibble: 3 × 3
#> # Groups: com [3]
#> com lci uci
#> <fct> <dbl> <dbl>
#> 1 hp 0.0134 0.165
#> 2 wt 0.126 0.382
#> 3 wt,hp 0.445 0.596
#> # A tibble: 3 × 3
#> # Groups: com [3]
#> com lci uci
#> <fct> <dbl> <dbl>
#> 1 hp 0.0259 0.184
#> 2 wt 0.105 0.343
#> 3 wt,hp 0.429 0.606
For additional details about on plotting with ggcommonality, see the ‘Advanced ggcommonality’ vignette:
vignette('advanced-ggcommonality', 'ggcommonality')
#> Warning: vignette 'advanced-ggcommonality' not foundYou can also read the help documentation ?ggcom(), or email
me.
Anderson, Cameron J, and Michael Schutz. 2022. “Exploring Historic Changes in Musical Communication: Deconstructing Emotional Cues in Preludes by Bach and Chopin.” Psychology of Music 50 (5): 1424–42.
Delle Grazie, Max, Cameron J Anderson, and Michael Schutz. 2025. “Breaking with Common Practice: Exploring Modernist Musical Emotion.” Psychology of Music, 03057356241296852.
Fox, John, and Sanford Weisberg. 2018. An r Companion to Applied Regression. Sage publications.
Nimon, Kim, Fred Oswald, and J. Kyle Roberts. 2023. Yhat: Interpreting Regression Effects. https://CRAN.R-project.org/package=yhat.