Visualizing the Summary Statistics of Anscombe’s Quartet
#I appreciated how straightforward this assignment was once I figured out how simple it can be to wrap the quartet of charts - it felt like a lightbulb moment in my R journey. It was also fun to explore the new `ggpubr` package that adds the linear regression equation and R² value directly onto the plots.
#load packages and import data
library(tidyverse)
library(ggpubr)
anscombes <- read.csv("./anscombes.csv")
#At first I started by splitting the dataset into separate frames,
#and plotting each one individually.
#I then realized that all of this was completely unnecessary for faceting into one output
#and there was a much easier way to do it.
#split into component datasets
setI <- anscombes[-c(1)] %>%
filter(dataset == "I")
setII <- anscombes[-c(1)] %>%
filter(dataset == "II")
setIII <- anscombes[-c(1)] %>%
filter(dataset == "III")
setIV <- anscombes[-c(1)] %>%
filter(dataset == "IV")
#create individual plots
plotI <- ggplot(setI, aes(x=x, y=y)) +
geom_point() +
xlim(c(2,18)) +
ylim(c(2,12)) +
theme_light()
plotII <- ggplot(setII, aes(x=x, y=y)) +
geom_point() +
xlim(c(2,18)) +
ylim(c(2,12)) +
theme_light()
plotIII <- ggplot(setIII, aes(x=x, y=y)) +
geom_point() +
xlim(c(2,18)) +
ylim(c(2,12)) +
theme_light()
plotIV <- ggplot(setIV, aes(x=x, y=y)) +
geom_point() +
xlim(c(2,18)) +
ylim(c(2,12)) +
theme_light()
#create faceted plots
#note: I ended up extending both of the axes to make more room for the annotations
facetplots <- ggplot(anscombes, aes(x = x, y = y)) +
geom_point(shape = 21, size = 4, color = "black", fill = "darkolivegreen") +
xlim(c(0,25)) +
ylim(c(0,20)) +
theme_light(base_size = 16) +
facet_wrap(~dataset)
#generate dataframe of summary statistics
#note: format() with the argument "nsmall" was my (perhaps inelegant) solution
#to forcing trailing zero decimals onto the whole numbers
summarystats.df <-
anscombes %>%
group_by(dataset) %>%
summarise(xmeans = mean(x), ymeans = mean(y), xSD = sd(x), ySD = sd(y)) %>%
mutate(label_xmeans = paste0("x mean: ", sep = "\n", format(xmeans, nsmall = 4)),
label_ymeans = paste0("y mean: ", sep = "\n", format(round(ymeans, digits = 4), nsmall = 2)),
label_xSD = paste0("x SD: ", sep = "\n", format(round(xSD, digits = 4), nsmall = 2)),
label_ySD = paste0("y SD: ", sep = "\n", format(round(ySD, digits = 4), nsmall = 2)))
#add summary stats for each dataset to faceted plots
facetplots.labels <-
facetplots +
geom_vline(data = summarystats.df, aes(xintercept = xmeans), color = "skyblue2", linewidth = 1) +
geom_label(data = summarystats.df, aes(x = xmeans, label = label_xmeans, y = 18), color = "skyblue2", label.size = 1) +
geom_hline(data = summarystats.df, aes(yintercept = ymeans), color = "deeppink4", linewidth = 1) +
geom_label(data = summarystats.df, aes(y = ymeans, label = label_ymeans, x = 23), color = "deeppink4", label.size = 1) +
geom_vline(data = summarystats.df, aes(xintercept = (xmeans-xSD)), color = "skyblue2", linewidth = 0.6, linetype = "dashed") +
geom_vline(data = summarystats.df, aes(xintercept = (xmeans+xSD)), color = "skyblue2", linewidth = 0.6, linetype = "dashed") +
geom_label(data = summarystats.df, aes(x = (xmeans-xSD), label = label_xSD, y = 14), color = "skyblue2", label.size = 1) +
#I initially had the SD lines at their numeric positions,
#but then realized they should be (SD units) AWAY from the mean line instead
geom_hline(data = summarystats.df, aes(yintercept = (ymeans-ySD)), color = "deeppink4", linewidth = 0.6, linetype = "dashed") +
geom_hline(data = summarystats.df, aes(yintercept = (ymeans+ySD)), color = "deeppink4", linewidth = 0.6, linetype = "dashed") +
geom_label(data = summarystats.df, aes(y = (ymeans-ySD), label = label_ySD, x = 18), color = "deeppink4", label.size = 1)
#add linear regression lines and r^2
facetplots.labels +
geom_smooth(method="lm", color = "black") +
stat_regline_equation(label.x = 18, label.y = 18) +
stat_cor(aes(label = after_stat(rr.label)), label.x = 18, label.y = 17)
