How to change the significant figures on R-squared in stat_regline_equation()?

Question

Here is a sample of my data:

df <- data.frame(surface_area = c(0.25, 1, 2.25, 4, 6.25),
             weight = c(0.8, 3.267, 7.733, 13.533, 21.8333))

The code for my plot and my plot:

    plot <- ggplot(data = df, aes(x = surface_area, y = weight)) + geom_point() +
  geom_smooth(method = 'lm',
              se = FALSE) +
  stat_regline_equation(aes(label = ..rr.label..),
                        label.x = 4, 
                        label.y = 5.8, 
                        size = 6) +
  stat_regline_equation(label.x = 4, 
                        label.y = 6.5, 
                        size = 6)

The line doesn't quite perfectly fit the data, but I am still getting a R-squared of 1. How do I increase the significant figures or decimal places of the R-squared to get a more accurate R-squared?

Many thanks!

https://stackoverflow.com/a/66183304/1412059 This is another reason why I would rather do this "manually", i.e., without ggpubr. — Roland, Mar 24 '23 at 12:28
There are other ways to add R^2 to the plot, see https://stackoverflow.com/q/7549694/680068 — zx8754, Mar 24 '23 at 12:32
Oh gosh, that is more than I think I can handle @Roland thank you tho! — dan_confused, Mar 24 '23 at 12:34
It is a known issue since 2020: https://github.com/kassambara/ggpubr/issues/312 — zx8754, Mar 24 '23 at 12:44

score 1 · Answer 1 · answered Mar 24 '23 at 12:55

Another option may be by first calculating the coefficient using lm with summary and after that change the coefficient that is plotted with gsub using ggplot_build to change the label of the layer that plots the coefficient. Here is some reproducible code:

library(ggplot2)
library(ggpubr)
model = lm(weight~surface_area, data = df)
r_squared = summary(model)$r.squared
r_squared
#> [1] 0.9995947
p = ggplot(data = df, aes(x = surface_area, y = weight)) + geom_point() +
  geom_smooth(method = 'lm',
              se = FALSE) +
  stat_regline_equation(aes(label = ..rr.label..),
                        label.x = 4, 
                        label.y = 5.8, 
                        size = 6) +
  stat_regline_equation(label.x = 4, 
                        label.y = 6.5, 
                        size = 6)

q <- ggplot_build(p)
q$data[[3]]$label = gsub("1", as.character(r_squared), q$data[[3]]$label)
q <- ggplot_gtable(q)
plot(q)

^{Created on 2023-03-24 with reprex v2.0.2}

If you want to round your coefficient you could use the round function on the calculate coefficient like this:

library(ggplot2)
library(ggpubr)
model = lm(weight~surface_area, data = df)
r_squared = summary(model)$r.squared
p = ggplot(data = df, aes(x = surface_area, y = weight)) + geom_point() +
  geom_smooth(method = 'lm',
              se = FALSE) +
  stat_regline_equation(aes(label = ..rr.label..),
                        label.x = 4, 
                        label.y = 5.8, 
                        size = 6) +
  stat_regline_equation(label.x = 4, 
                        label.y = 6.5, 
                        size = 6)

q <- ggplot_build(p)
q$data[[3]]$label = gsub("1", as.character(round(r_squared, 4)), q$data[[3]]$label)
q <- ggplot_gtable(q)
plot(q)

^{Created on 2023-03-24 with reprex v2.0.2}

How to change the significant figures on R-squared in stat_regline_equation()?

1 Answers1