Factor order within faceted dotplot using ggplot2

Question

I\'m trying to change the plotting order within facets of a faceted dotplot in ggplot2, but I can\'t get it to work. Here\'s my melted dataset:

> London.m         


        

Answer 1

This is the best I can do with qplot. Not exactly what you asked for but closer. OOOPs I see you already figured that out.

q <- qplot(x = count, y = reorder(country, count), data = London.melt, geom = "point", facets = medal.type ~.)

Here's a dput version so others can improve:

dput(London.melt)
structure(list(country = structure(c(9L, 6L, 3L, 1L, 7L, 4L, 
5L, 8L, 2L, 10L, 9L, 6L, 3L, 1L, 7L, 4L, 5L, 8L, 2L, 10L, 9L, 
6L, 3L, 1L, 7L, 4L, 5L, 8L, 2L, 10L), .Label = c("Australia", 
"China", "France", "Germany", "Great Britain & N. Ireland", "Italy", 
"Japan", "Russian Federation", "South Korea", "United States"
), class = "factor"), medal.type = structure(c(2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), .Label = c("bronze", 
"gold", "silver"), class = "factor"), count = c(13L, 8L, 11L, 
7L, 7L, 11L, 29L, 24L, 38L, 46L, 8L, 9L, 11L, 16L, 14L, 19L, 
17L, 26L, 27L, 29L, 7L, 11L, 12L, 12L, 17L, 14L, 19L, 32L, 23L, 
29L)), .Names = c("country", "medal.type", "count"), class = "data.frame", row.names = c("1", 
"2", "3", "4", "5", "6", "7", "8", "9", "10", "11", "12", "13", 
"14", "15", "16", "17", "18", "19", "20", "21", "22", "23", "24", 
"25", "26", "27", "28", "29", "30"))

Answer 2

Here a solution using paste, free scales and some relabeling

library(ggplot2)
London.melt$medal.type<-factor(London.melt$medal.type, levels = c("gold","silver","bronze"))
# Make every country unique
London.melt$country_l <- with(London.melt, paste(country, medal.type, sep = "_"))
#Reorder the unique countrys
q <- qplot(x = count, y = reorder(country_l, count), data = London.melt, geom = "point") +   facet_grid(medal.type ~., scales = "free_y")
# Rename the countries using the original names
q + scale_y_discrete("Country", breaks = London.melt$country_l, label = London.melt$country)

Answer 3

This is obviously quite late, and some of what I'm doing may have not been around 6 years ago, but I came across this question while doing a similar task. I'm always reluctant to set tick labels with a vector—it feels safer to use a function that can operate on the original labels.

To do that, I'm creating a factor ID column based on the country and the medal, with some delimiter character that doesn't already appear in either of those columns—in this case, _ works. Then with forcats::fct_reorder, I can order that column by count. The last few levels of this column are below, and should correspond to the country + medal combinations with the highest counts.

library(tidyverse)

London_ordered <- London.melt %>%
  mutate(id = paste(country, medal.type, sep = "_") %>%
           as_factor() %>%
           fct_reorder(count, .fun = min))

levels(London_ordered$id) %>% tail()
#> [1] "Great Britain & N. Ireland_gold" "United States_silver"           
#> [3] "United States_bronze"            "Russian Federation_bronze"      
#> [5] "China_gold"                      "United States_gold"

Then use this ID as your y-axis. On its own, you'd then have very long labels that include the medal type. Because of the unique delimiter, you can write an inline function for the y-axis labels that will remove the delimiter and any word characters that come after it, leaving you with just the countries. Moving the facet specification to a facet_wrap function lets you then set the free y-scale.

qplot(x = count, y = id, data = London_ordered, geom = "point") +
  scale_y_discrete(labels = function(x) str_remove(x, "_\\w+$")) +
  facet_wrap(~ medal.type, scales = "free_y", ncol = 1)