Plot2WayANOVA: Plot a 2 Way ANOVA using dplyr and ggplot2

Description

Takes a formula and a dataframe as input, conducts an analysis of variance prints the results (AOV summary table, table of overall model information and table of means) then uses ggplot2 to plot an interaction graph (line or bar) . Also uses Brown-Forsythe test for homogeneity of variance. Users can also choose to save the plot out as a png file.

Usage

Plot2WayANOVA(formula, 
               dataframe = NULL, 
               confidence=.95,
               plottype = "line", 
               xlab = NULL, 
               ylab = NULL, 
               title = NULL,
               subtitle = NULL, 
               interact.line.size = 2, 
               ci.line.size = 1, 
               mean.label = FALSE,
               mean.ci = TRUE, 
               mean.size = 4, 
               mean.shape = 23, 
               mean.color = "darkred",
               mean.label.size = 3, 
               mean.label.color = "black", 
               offset.style = "none",
               overlay.type = NULL,
               posthoc.method = "scheffe",
               show.dots = FALSE,
               PlotSave = FALSE)

Arguments

formula

a formula with a numeric dependent (outcome) variable, and two independent (predictor) variables e.g. mpg ~ am * vs. The independent variables are coerced to factors (with warning) if possible.

dataframe

a dataframe or an object that can be coerced to a dataframe

confidence

what confidence level for confidence intervals

plottype

bar or line (quoted)

xlab, ylab

Labels for `x` and `y` axis variables. If `NULL` (default), variable names for `x` and `y` will be used.

title

The text for the plot title. A generic default is provided.

subtitle

The text for the plot subtitle. If `NULL` (default), key model information is provided as a subtitle.

interact.line.size

Line size for the line connecting the group means (Default: `2`).

ci.line.size

Line size for the confidence interval bracketing the group means (Default: `1`).

mean.label

Logical that decides whether the value of the group mean is to be displayed (Default: `FALSE`).

mean.ci

Logical that decides whether the confidence interval for group means is to be displayed (Default: `TRUE`).

mean.size

Point size for the data point corresponding to mean (Default: `4`).

mean.shape

Shape of the plot symbol for the mean (Default: `23` which is a diamond).

mean.color

Color for the data point corresponding to mean (Default: `"darkred"`).

mean.label.size, mean.label.color

Aesthetics for the label displaying mean. Defaults: `3`, `"black"`, respectively.

offset.style

A character string (e.g., `"wide"` or `"narrow"`, or `"none"`) which controls whether items are offset from the centerline for clarity. Useful when you want to add individual datapoints or confdence interval lines overlap. (Default: `"none"`).

overlay.type

A character string (e.g., `"box"` or `"violin"`), if you wish to overlay that information on factor1

posthoc.method

A character string, one of "hsd", "bonf", "lsd", "scheffe", "newmankeuls", defining the method for the pairwise comparisons. (Default: `"scheffe"`).

show.dots

Logical that decides whether the individual data points are displayed (Default: `FALSE`).

PlotSave

a logical indicating whether the user wants to save the plot as a png file

Value

A list with 5 elements which is returned invisibly. These items are always sent to the console for display but for user convenience the function also returns a named list with the following items in case the user desires to save them or further process them - $ANOVATable,$ModelSummary, $MeansTable, $PosthocTable, $BFTest, and $SWTest. The plot is always sent to the default plot device

Details

Details about how the function works in order of steps taken.

Some basic error checking to ensure a valid formula and dataframe. Only accepts fully *crossed* formula to check for interaction term
Ensure the dependent (outcome) variable is numeric and that the two independent (predictor) variables are or can be coerced to factors -- user warned on the console
Remove missing cases -- user warned on the console
Calculate a summarized table of means, sds, standard errors of the means, confidence intervals, and group sizes.
Use aov function to execute an Analysis of Variance (ANOVA)
Use anova_stats to calculate eta squared and omega squared values per factor. If the design is unbalanced warn the user and use Type II sums of squares
Produce a standard ANOVA table with additional columns
Use the PostHocTest for producing a table of post hoc comparisons for all effects that were significant
Use the leveneTest for testing Homogeneity of Variance assumption with Brown-Forsythe
Use the PostHocTest for conducting post hoc tests for effects that were significant
Use the shapiro.test for testing normality assumption with Shapiro-Wilk
Use ggplot2 to plot an interaction plot of the type the user specified.

The defaults are deliberately constructed to emphasize the nature of the interaction rather than focusing on distributions. So while a violin plot of the first factor by level is displayed along with dots for individual data points shaded by the second factor, the emphasis is on the interaction lines.

References

: ANOVA: Delacre, Leys, Mora, & Lakens, *PsyArXiv*, 2018

Examples

Run this code

# NOT RUN {
Plot2WayANOVA(mpg ~ am * cyl, mtcars, plottype = "line")
Plot2WayANOVA(mpg ~ am * cyl, 
              mtcars, 
              plottype = "line", 
              overlay.type = "box", 
              mean.label = TRUE)
Plot2WayANOVA(mpg ~ am * vs, mtcars, confidence = .99)

# Create a new dataset
library(dplyr)
library(ggplot2)
library(stringi)
newmpg <- mpg %>% 
          filter(cyl != 5) %>% 
          mutate(am = stringi::stri_extract(trans, regex = "auto|manual"))
Plot2WayANOVA(formula = hwy ~ am * cyl,
              dataframe = newmpg,
              ylab = "Highway mileage",
              xlab = "Transmission type",
              plottype = "line",
              offset.style = "wide",
              overlay.type = "box",
              mean.label = TRUE, 
              mean.shape = 20, 
              mean.size = 5, 
              mean.label.size = 5,
              show.dots = TRUE)

# }

Run the code above in your browser using DataLab