Estimation Statistics with Confidence Intervals

A collection of functions and 'jamovi' module for the estimation approach to inferential statistics, the approach which emphasizes effect sizes, interval estimates, and meta-analysis. Nearly all functions are based on 'statpsych' and 'metafor'. This package is still under active development, and breaking changes are likely, especially with the plot and hypothesis test functions. Data sets are included for all examples from Cumming & Calin-Jageman (2024) .

esci: Estimation Statistics with Confidence Intervals

esci provides student-friendly tools for estimation statistics:

• effect sizes with confidence intervals for many research designs
• meta-analysis
• visualizations emphasizing effect sizes and uncertainty
• strong hypothesis testing with interval nulls

esci is both an R package and a module in jamovi. If you're looking for the R package, stay here. If you want esci in jamovi, download and install jamovi and then use the module library to add esci.

Leave comments, bug reports, suggestions, and questions about esci here

esci is still under development; expect breaking changes in the future especially for the visualization functions. If you need production-ready estimation, turn to statpsych

esci is built on top of statpsych and metafor. That is, almost all of the statistical calculations are passed off to these packages. The only exception is for confidence intervals for Cohen's d (see documentation). Why does esci exist, then?

• To provide a design-based approach; each function in esci is for one type of research design (e.g. two groups with a continuous variable); it provides all the effect sizes relevant to that design in one convenient function (e.g. mean difference, cohen's d, median difference, ratio of means, ratio of medians).
• To make visualization easier; esci provides visualizations that emphasize effect sizes and uncertainty
• To integrate with GUIs for students; esci integrates into jamovi and (very soon) JASP.
• To support student learning, from intro to advanced levels; esci is extensively used in the recently released second edition of our intro textbook (https://thenewstatistics.com/itns/) and should be ideal for use in higher level and graduate courses.

The visualizations produced by esci are exquisite in a large part because of the lovely ggdist package by Matthew Kay.

Installation

esci is availableon CRAN; you can install with:

install.packages("esci")

Or, get the stable branch directly from github

# install.packages("devtools")
devtools::install_github('rcalinjageman/esci')

Or, try out the development branch:

# install.packages("devtools")
devtools::install_github('rcalinjageman/esci',  branch = "development")

Roadmap

• ~Finish writing documentation and tests~
• Review all functions for consistency of parameter names and returned object names
• Rewrite visualization functions completely to remove clunky approaches to the difference axis and other issues
• ~Complete JASP integration~
• Rewrite jamovi integration
• Add prediction intervals for basic designs
• Repeated measures with 1 IV and multiple groups
• Fully within-subjects 2x2 design
• Arbitrarily complex designs

Example

library(esci)

data("data_kardas_expt_4")
estimate <- estimate_mdiff_two(data_kardas_expt_4, Prediction, Exposure)
plot_mdiff(estimate)