Ordering facets by size instead of alphabetical (time series data)

[Piranha]

Hello,

This is part 2 of an earlier question that I had posted (part 1: Facet chart in a Shiny web app with lots of plots)

I have a shiny app that creates plots based on some grouping criteria. I want the individual facets/plots to be sorted by size. In the case of the gapminder example, by gpdPercap. I know that there are several posts about this (example: Ordering categories within ggplot2 facets | R-bloggers), but I am having a really hard time applying it to my time series example.

The current code and facet chart is shown below. Any help would be much appreciated!!

library(shiny)
library(tidyverse)
library(gapminder)

### Pre-process Data

## functions
pct_change <- function(x) {
  ((x - lag(x,1)) / lag (x,1)) * 100
}



## data frame to be used in shiny app
df <- gapminder %>% 
  select(country, year, gdpPercap) %>% 
  group_by(country) %>% 
  mutate(gdpPercap_change = pct_change(gdpPercap)) %>% 
  mutate(gdpPercap_change = round(gdpPercap_change, 2)) %>%  
  mutate(growth_group = case_when(year == 2007 & gdpPercap_change > 20                        ~ "high growth",
                                  year == 2007 & gdpPercap_change < 20 & gdpPercap_change > 0 ~ "moderate growth",
                                  year == 2007 & gdpPercap_change < 0                         ~ "declining",
                                  TRUE                                                        ~ as.character(NA)
  )) %>% 
  fill(growth_group, .direction = "up") %>% 
  na.omit()
head(df)


### UI portion
ui <- fluidPage(
  titlePanel("Gapminder: Dashboard"),
  
  sidebarLayout(
    sidebarPanel(
      selectInput(inputId="group", label="Choose a Growth Group", 
                  choices=unique(df$growth_group), selected="declining")
    ),
    mainPanel(
      plotOutput("chart_1")
    )
  )
)


### Server portion
server <- function(input, output) {
  
  output$chart_1 <- renderPlot({
    chart <- df %>% 
      filter(growth_group == input$group) %>% 
      ungroup() %>% 
      ggplot(aes(x=year, y=gdpPercap_change)) +
      geom_line() +
      geom_hline(yintercept = 0) +
      facet_wrap(~ country, scales = "free_y", ncol = 5)
    chart
  })
}

### run shiny app
shinyApp(ui = ui, server = server)

[martin.R]

You just need to define country as a factor ordered by your desired criteria, e.g. using

Reorder factor levels by sorting along another variable — fct_reorder

Time series data or not does not affect this as long as you can define the ordering properly (e.g. last or mean gdpPercap).

[mattwarkentin]

As mentioned by @martin.R, reordering the country by some other measure should do the trick:

library(forcats)

server <- function(input, output) {
  
  output$chart_1 <- renderPlot({
    chart <- df %>% 
      filter(growth_group == input$group) %>% 
      ungroup() %>% 
      mutate(country = fct_reorder(country, gdpPercap)) %>% 
      ggplot(aes(x=year, y=gdpPercap_change)) +
      geom_line() +
      geom_hline(yintercept = 0) +
      facet_wrap(~ country, scales = "free_y", ncol = 5)
    chart
  })
}

NOTE: By default, fct_reorder() orders by median() of the x variable (gdpPercap), you can specify mean, max, etc. by specifying a .fun. For example:

server <- function(input, output) {
  
  output$chart_1 <- renderPlot({
    chart <- df %>% 
      filter(growth_group == input$group) %>% 
      ungroup() %>% 
      mutate(country = fct_reorder(country, gdpPercap, .fun = mean)) %>% 
      ggplot(aes(x=year, y=gdpPercap_change)) +
      geom_line() +
      geom_hline(yintercept = 0) +
      facet_wrap(~ country, scales = "free_y", ncol = 5)
    chart
  })
}

[Piranha]

Ah, many thanks! This solved my question perfectly!

Need to read up more details on forcats. Not sure why I didn't look into that until now.