Using geom_map with transition_time gets Error: Error in insert_points(polygon$x, polygon$y, splits, n)

I'm trying to show the growth of COVID cases in New York state

This code gets the plot I want but without the animation or aspect of time.

Full error:

Error in insert_points(polygon$x, polygon$y, splits, n):
Not compatible with requested type: [type=NULL; target=double].

library(ggplot2)
library(gganimate)
library(transformer)
library(tidyverse)
county_map = map_data("county", region = "New York")

county_map$region = county_map$subregion

covidCounties = read.csv("https://raw.githubusercontent.com/nytimes/covid-19-data/master/us-counties.csv", header = T)
covidCounties = covidCounties %>%
  mutate(date = as.Date(date)) %>%
  filter(state == "New York") %>%
  arrange(date)%>%
  group_by(county) %>%
  mutate(county = tolower(county)) %>%
  mutate(newCases = diff(c(0, cases))) %>%
  mutate(newDeaths = diff(c(0, deaths))) %>%
  ungroup() %>%
  select(date, state, county, cases, newCases, deaths)

covidCountyMap = covidCounties %>%
  ggplot(aes(
    map_id = county,
    fill = newCases,
    group = county
  ))+
  geom_map(
    map = county_map,
    color = "black"
  )+
  expand_limits(x = county_map$long, y = county_map$lat)+
  scale_fill_gradientn(colors = c("green", "yellow", "red"), breaks = c(0, 100, 500))+
  labs(
    title = "New cases over time in New York State",
    subtitle = "{frame_time}"
  )

covidCountyMap

covidCountyMap+
  transition_time(date)

Solution

You need to tell {gganimate} what polygons to transition to one another. It won't be able to guess that for you. In other words, you need to add a group identifier to each transition state (meaning each county by date).

I filtered to only one state because the reprex on the entire data kept crashing. I have transformed to a log scale for your counts, in order to represent the data range better. (there are a few negative values, therefore the warning)

library(tidyverse)
library(gganimate)

county_map = map_data("county", region = "New York")

county_map$region = county_map$subregion

## I'd advise to create a separate data frame for your raw data, and not overwrite it
covidCounties_raw = read.csv("https://raw.githubusercontent.com/nytimes/covid-19-data/master/us-counties.csv", header = T)

covidCounties <- covidCounties_raw %>%
  mutate(date = as.Date(date)) %>%
  filter(state == "New York") %>%
  arrange(date) %>%
  group_by(county) %>%
  mutate(county = tolower(county)) %>%
  mutate(newCases = diff(c(0, cases))) %>%
  mutate(newDeaths = diff(c(0, deaths))) %>%
  ungroup() %>%
  select(date, state, county, cases, newCases, deaths) %>%
  ## this is the main trick
  group_by(date, county) %>%
  mutate(id = cur_group_id()) %>%
  ungroup() %>%
  ## I'm filtering for only one county because the reprex took too long with the entire data
  filter(county == "nassau")

covidCountyMap <- covidCounties %>%
  ggplot(aes(
    map_id = county,
    fill = newCases,
## use the group identifier for your grouping
    group = id
  )) +
  geom_map(
    map = county_map,
    color = "black"
  ) +
  expand_limits(x = county_map$long, y = county_map$lat) +
  scale_fill_gradientn(colors = c("green", "yellow", "red"),
## log transformed scale
                       trans = "log") +
  labs(
    title = "New cases over time in New York State",
    subtitle = "{frame_time}"
  )

anim <- covidCountyMap +
  transition_time(date)

## have slightly reduced the frame rate to make it slightly faster
animate(anim, fps = 5, nframes = 50)
#> Warning: Transformation introduced infinite values in discrete y-axis

^{Created on 2021-11-30 by the reprex package (v2.0.1)}