In DJAnderson07/esvis: Visualization and Estimation of Effect Sizes

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "README-"
)

esvis

R Package for effect size visualization and estimation.

This package is designed to help you very quickly estimate and visualize distributional differences by categorical factors (e.g., the effect of treatment by gender and income category). Emphasis is placed on evaluating distributional differences across the entirety of the scale, rather than only by measures of central tendency (e.g., means).

Installation

Install directly from CRAN with

install.packages("esvis")

Or the development version from GitHub with:

# install.packages("devtools")
devtools::install_github("datalorax/esvis")

Plotting methods

There are three primary data visualizations: (a) binned effect size plots, (b) probability-probability plots, and (c) empirical cumulative distribution functions. All plots use the ggplot2 package and are fully manipulable after creation using standard ggplot commands (e.g., changing the theme, labels, etc.). These plots were all produced by first running library(ggplot2); theme_set(theme_minimal()) to produce the plots with the minimal theme, but no theme structure is imposed on any of the plots.

Binned ES Plot

At present, the binned effect size plot can only be produced with Cohen's d, although future development will allow the user to select the type of effect size. The binned effect size plot splits the distribution into quantiles specified by the user (defaults to lower, middle, and upper thirds), calculates the mean difference between groups within each quantile bin, and produces an effect size for each bin by dividing by the overall pooled standard deviation (i.e., not by quantile). For example

library(ggplot2)
library(esvis)
theme_set(theme_minimal())

library(esvis)
binned_plot(benchmarks, math ~ ell)

Note that in this plot one can clearly see that the magnitude of the differences between the groups depends upon scale location, as evidence by the reversal of the effect (negative to positive) for the Non-ELL (non-English Language Learners) group. We could also change the reference group, change the level of quantile binning, and evaluate the effect within other factors. For example, we can look by season eligibility for free or reduced price lunch, with quantiles binning, and non-ELL students as the reference group with

binned_plot(benchmarks, 
            math ~ ell + frl + season, 
            ref_group = "Non-ELL",
            qtile_groups = 5)

The ref_group argument can also supplied as a formula.

PP Plots

Probability-probability plot can be produced with a call to pp_plot and an equivalent argument structure. In this case, we're visualizing the difference in reading achievement by race/ethnicity by season.

pp_plot(benchmarks, reading ~ ethnicity + season)

Essentially, the empirical cummulative distribution function (ECDF) for the reference group (by default, the highest performing group) is mapped against the ECDF for each corresponding group. The magnitude of the achievement gap is then displayed by the distance from the diagonal reference line, representing, essentially, the ECDF for the reference group.

By default, the area under the curve is shaded, which itself is an effect-size like measure, but this is also manipulable.

ECDF Plot

Finally, the ecdf_plot function essentially dresses up the base plot.ecdf function, but also adds some nice referencing features through additional, optional arguments. Below, I have included the optional hor_ref = TRUE argument such that horizontal reference lines appear, relative to the cuts provided.

ecdf_plot(benchmarks, math ~ season, 
    cuts = c(190, 200, 215))

These are the curves that go into the PP-Plot, but occasionally can be useful on their own.

Estimation Methods

Compute effect sizes for all possible pairwise comparisons.

coh_d(benchmarks, math ~ season + frl)

Or specify a reference group. In this case, I've used the formula-based interface, but a string vector specifying the specific reference group could also be supplied.

coh_d(benchmarks, 
      math ~ season + frl, 
      ref_group = ~Fall + `Non-FRL`)

Notice that the reference to Non-FRL is wrapped in back-ticks, which should be used anytime there are spaces or other non-standard characters.

Other effect sizes are estimated equivalently. For example, compute V (Ho, 2009) can be estimated with

v(benchmarks, 
  math ~ season + frl, 
  ref_group = ~Fall + `Non-FRL`)

or AUC with

auc(benchmarks, 
    math ~ season + frl, 
    ref_group = ~Fall + `Non-FRL`)

DJAnderson07/esvis documentation built on May 5, 2020, 3:20 p.m.