paired outcomes Stat

[EvaMaeRey]

A good grouping problem...

library(tidyverse)
compute_panel_pairs <- function(data, scales){
  
  data %>% 
    group_by(grouping) %>% 
    arrange(x) %>% 
    mutate(change = y - lag(y)  ) %>% 
    fill(change, .direction = "up")  
  
}

StatPairs <- ggproto("StatPairs", Stat,
                     compute_panel = compute_panel_pairs,
                     default_aes = aes(color = after_stat(change)))


gapminder::gapminder %>% 
  filter(year %in% c(1992, 2002)) %>% 
  ggplot() + 
  aes(x = year, y = gdpPercap, grouping = country) + 
  geom_point(stat = StatPairs) +
  geom_path(stat = StatPairs)

last_plot() + 
  aes(x = as.factor(years))
#> Error in `geom_point()`:
#> ! Problem while computing aesthetics.
#> ℹ Error occurred in the 1st layer.
#> Caused by error in `unique.default()`:
#> ! unique() applies only to vectors

^{Created on 2024-08-23 with reprex v2.1.0}

[danilyanedo7]

The error is being caused because you're trying to add an aesthetic aes(x = as.factor(years) to the last_plot() function, but the variable years doesn’t exist in your dataset. Also, adding an aesthetic to an existing plot doesn't modify the data or aesthetics in the layers that have already been created.

# Setup
library(tidyverse)

Custom compute panel function to calculate the change

compute_panel_pairs <- function(data, scales){
  data %>%
    group_by(group) %>%  # Group by the 'group' aesthetic which refers to 'country'
    arrange(x) %>%
    mutate(change = y - lag(y)) %>% 
    fill(change, .direction = "up")
}

Define the custom stat

StatPairs <- ggproto("StatPairs", Stat,
                     compute_panel = compute_panel_pairs,
                     default_aes = aes(color = after_stat(change), group = country))

Use the custom stat in your plot

gapminder::gapminder %>%
  filter(year %in% c(1992, 2002)) %>%
  ggplot() + 
  aes(x = as.factor(year), y = gdpPercap, group = country) +  # Use 'as.factor(year)'
  geom_point(stat = StatPairs) +
  geom_path(stat = StatPairs)

[EvaMaeRey]

Thank you so much for pointing out the 'years' problem! That could have had me stuck forever. years is a function in the tidyverse, and that's what the error message was about. Found a function, where a vector was expected - but I attributed the problem to how I wrote the Stat.... 🤦‍♀️

I rewrote the compute a little, with group aesthetic, which I'm less confident about, but it does seem to work, and with the factor version of the variable. I should probably stress test a little more.

Using aes outside of the ggplot() or geom_*() parens is not advertized, and post-geom declarations are probably even less common, but it can be pretty useful in testing/demoing behavior. Some examples below. Thank you again!

library(tidyverse)
compute_panel_pairs <- function(data, scales){
  
  data %>% 
    group_by(group) %>% 
    arrange(x) %>% 
    mutate(change = y - lag(y)  ) %>% 
    fill(change, .direction = "up")  
  
}

StatPairs <- ggproto("StatPairs", Stat,
                     compute_panel = compute_panel_pairs,
                     default_aes = aes(color = after_stat(change)))

gapminder::gapminder %>% 
  filter(year %in% c(1992, 2002)) %>% 
  ggplot() + 
  aes(x = year, 
      y = gdpPercap, 
      group = country) + 
  geom_point(stat = StatPairs) +
  geom_path(stat = StatPairs)

last_plot() + 
  aes(x = as.factor(year))

last_plot() + 
  aes(color = sign(after_stat(change)))

last_plot() + 
  aes(color = I("grey"))

^{Created on 2024-09-05 with reprex v2.1.0}