https://github.com/dcousin3/anopa
Analysis of proportions using Anscombe transform
https://github.com/dcousin3/anopa
error-bars proportions r statistical-testing statistics summary-statistics
Last synced: 3 months ago
JSON representation
Analysis of proportions using Anscombe transform
- Host: GitHub
- URL: https://github.com/dcousin3/anopa
- Owner: dcousin3
- Created: 2024-03-16T23:56:47.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2024-03-21T17:14:28.000Z (about 1 year ago)
- Last Synced: 2024-04-25T20:45:53.092Z (about 1 year ago)
- Topics: error-bars, proportions, r, statistical-testing, statistics, summary-statistics
- Language: R
- Homepage: https://dcousin3.github.io/ANOPA/
- Size: 3.55 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.Rmd
Awesome Lists containing this project
README
---
output: github_document
bibliography: "inst/REFERENCES.bib"
csl: "inst/apa-6th.csl"
---
# ANOPA: Analysis of Proportions using Anscombe transform
[](https://cran.r-project.org/package=ANOPA)
```{r, echo = FALSE, message = FALSE, results = 'hide', warning = FALSE}
cat("this will be hidden; used for general initializations.\n")
library(ANOPA)
options("ANOPA.feedback" = "none") # shut down all information
```
The library `ANOPA` provides easy-to-use tools to analyze proportions .
With it, you can examine
if proportions are significantly different (_show an effect_). In the
case where there is more than one factor, you can also test if the
interaction(s) are significant. You
can also test simple effects (a.k.a. _expected marginal_ analysis),
as well as post-hoc tests (using Tukey's _Honestly Significant Difference_ test HSD).
Finally, you can assess differences based on orthogonal contrasts.
You can consult @lc23 for details.
ANOPA also comes (a) with tools to make a plot of the proportions along
with 95% confidence intervals [these intervals are adjusted for pair-
wise comparisons; @cgh21]; (b) with tools to compute statistical power given
some _a priori_ expected proportions or sample size to reach a certain
statistical power; (c) to generate random proportions if you wish to perform
Monte Carlo simulations on proportions.
In sum, everything you need to analyse proportions!
The main function is `anopa()` which returns an omnibus analysis of the
proportions for the factors given. For example, if you have a data frame
`ArticleExample2` which contains a column called `s` where the
number of successes per group are stored,
and a column called `n` where the group sizes are stored, then the following
performs an analysis of proportions as a function of the groups based on the
columns `SES` and `MofDiagnostic`:
```{r, message=FALSE, warning=FALSE, echo=TRUE, eval=TRUE}
w <- anopa( {s; n} ~ SES * MofDiagnostic, ArticleExample2 )
summary(w)
```
As the results suggest (consult the first three columns), there is
a main effect of the factor SES (F(2, inf) = 6.395, p = .002).
A plot of the proportions can be obtained easily with
```{r, fig.alt="anopa plot of proportions with confidence intervals", message=FALSE, warning=FALSE, fig.width=5.25, fig.height=3}
anopaPlot(w)
```
or just the main effect figure with
```{r, fig.alt = "anopa plot of proportions with confidence intervals", message=FALSE, warning=FALSE, fig.width=4, fig.height=3}
anopaPlot(w, ~ SES)
```
If the interaction had been significant, simple effects can be analyzed from the _expected marginal
frequencies_ with `e <- emProportions(w, ~ SES | MofDiagnostic )`.
Follow-up analyses include contrasts examinations with `contrastProportions()`; finally,
post-hoc pairwise comparisons can be obtained with `posthocProportions()`.
Prior to running an experiment, you might consider some statistical power planning
on proportions using ``anopaPower2N()`` or
``anopaN2Power()`` as long as you can anticipate the expected proportions. A
convenient effect size, the f-square and eta-square can be obtained with `anopaProp2fsq()`.
Finally, `toCompiled()`, `toLong()` and `toWide()`
can be used to present the proportion in other formats.
# Installation
The official **CRAN** version can be installed with
```{r, echo = TRUE, eval = FALSE}
install.packages("ANOPA")
library(ANOPA)
```
The development version `r packageVersion("ANOPA")` can be accessed through GitHub:
```{r, echo = TRUE, eval = FALSE}
devtools::install_github("dcousin3/ANOPA")
library(ANOPA)
```
Note that the package `ANOPA` is named using UPPERCASE letters whereas the main function
`anopa()` is written using lowercase letters.
The library is loaded with
```{r, echo = TRUE, eval = FALSE, results = FALSE}
library(ANOPA)
```
# In sum
As seen, the library `ANOPA` makes it easy to analyze proportions using the
same general vocabulary found in ANOVAs.
The complete documentation is available on this
[site](https://dcousin3.github.io/ANOPA/).
A general introduction to the `ANOPA` framework underlying this
library can be found at @lc23.
# References
\insertAllCited{}