stat_binhex hexagons appear extremely stretched if all data points fall in a single bin on y dimension

[ail24]

If all data points are in a single bin on the y axis (e.g. if they are all equal to 0) the hexagons appear extremely stretched vertically.

Minimal reprex:

data <- tibble(x = rnorm(1000, 20, 20), y = rep(0,1000))

data %>%
  ggplot(aes(x = x, y = y))+
  stat_binhex(bins = 100)+
  scale_x_continuous(limits = c(-100,100))+
  scale_y_continuous(limits = c(-100,100))

[smouksassi]

for stat_binhex the provided limits will be used for computing the ybns

ybnds <- hex_bounds(y, binwidth[2])

what output do you expect? you might want stat count where you are simply counting how many y at each unique x position

[clauswilke]

Something is definitely wrong when the variation in the data is less than the bin width in that direction. Even if binwidth is provided explicitly the plot fails.

library(tidyverse)

set.seed(1234)
data <- tibble(
  x = rnorm(1000, 20, 20),
  y1 = rnorm(1000, 0, 2),
  y2 = rnorm(1000, 0, .2)
)

data %>%
  ggplot(aes(x = x, y = y1)) +
  stat_binhex() +
  scale_x_continuous(limits = c(-100, 100)) +
  scale_y_continuous(limits = c(-100, 100))

data %>%
  ggplot(aes(x = x, y = y2)) +
  stat_binhex() +
  scale_x_continuous(limits = c(-100, 100)) +
  scale_y_continuous(limits = c(-100, 100))

data %>%
  ggplot(aes(x = x, y = y1)) +
  stat_binhex(binwidth = c(5, 5)) +
  scale_x_continuous(limits = c(-100, 100)) +
  scale_y_continuous(limits = c(-100, 100))

data %>%
  ggplot(aes(x = x, y = y2)) +
  stat_binhex(binwidth = c(5, 5)) +
  scale_x_continuous(limits = c(-100, 100)) +
  scale_y_continuous(limits = c(-100, 100))

# without limits

data %>%
  ggplot(aes(x = x, y = y2)) +
  stat_binhex()

data %>%
  ggplot(aes(x = x, y = y2)) +
  stat_binhex(binwidth = c(5, 5))

^{Created on 2021-05-05 by the reprex package (v1.0.0)}

[clauswilke]

My previous post wasn't entirely correct. Manually setting the bin height does work when limits aren't set. I just had picked a poor value in my previous example.

library(tidyverse)

set.seed(1234)
data <- tibble(
  x = rnorm(1000, 20, 20),
  y1 = rnorm(1000, 0, 2),
  y2 = rnorm(1000, 0, .2)
)

data %>%
  ggplot(aes(x = x, y = y2)) +
  stat_binhex(binwidth = c(5, 0.08))

^{Created on 2021-05-05 by the reprex package (v1.0.0)}

[thomasp85]

Potentially related #4580 ?

[thomasp85]

Fixed by #4671