In mathesong/pwrcontour: Power Contours And Tables For Different Hypothetical Effect Sizes

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

pwrcontour

The goal of pwrcontour is to present power analyses in which it is possible to evaluate power for numerous hypothetical effect sizes at once. This makes it easier to communicate the fact that power is a calibration for different potential contingencies of the true size of the effect of interest. This is based on the tables and figures created by Richard Morey in the jamovi jpower module.

This package is still in alpha, and more features will likely be added at a later point.

Installation

You can install the released version of pwrcontour from Github with:

remotes::install_github("mathesong/pwrcontour")

Example

The main functions are simply for plotting power contours, or examining power tables, based around the syntax of the functions in the pwr package.

Correlations

library(pwrcontour)

pwr.r.test.contour(nmin = 5, nmax = 100, rmin = 0.1, rmax = 0.7)
pwr.r.test.table(n = 25)

T-tests

pwr.t.test.contour(nmin = 5, nmax = 100, esmin = 0.1, esmax = 1)
pwr.t.test.table(n = 25)

This package also makes it possible to present these figures in the raw units of analysis. For example, if we know that a 10% change is equivalent to a Cohen's d of 1, we can incorporate this into the figure.

pwr.t.test.contour(nmin = 5, nmax = 100, esmin = 1, esmax = 10, 
                   d_unitconversion = 10, d_units = "%")
pwr.t.test.table(25, d_unitconversion = 10, d_units = "%")

Proportion tests

Then there's also proportion tests. I decided to discard the Cohen's h, as this feels like it complicates things unnecessarily. So we set up a proportion to compare to instead, called propcomp

pwr.prop.test.contour(nmin = 5, nmax = 100, propmin = 0.01, propmax = 0.99, 
                      propcomp = 0.75)
pwr.prop.test.table(n=75, propcomp=0.75)

mathesong/pwrcontour documentation built on Jan. 1, 2021, 9:19 a.m.