Skip to main content

数据科学中的 R 语言

第 33 章 ggplot2之让你的数据骚动起来

这节课，我们讲如何让我们的图动起来。（因为渲染需要花费很长时间，所以文档中的动图代码都没有执行。）

33.1 为什么要使用动图

  • 改进了图形在时间上和空间上的重新定位
  • 传递更多信息
  • 引人注意

33.2 gganimate宏包

动图可以将其理解为多张静态图堆在一起，当然不是随意的堆放，而是按照一定的规则，比如按照时间的顺序，或者类别的顺序。一般而言，动图制作包括两个步骤: 静态图制作及图形组装。静态图制作，前面几章我们讲过主要用ggplot2宏包实现；对于图形组装，需要用到今天我们要讲Thomas Lin Pedersen的gganimate宏包，来自同一工厂的产品，用起来自然是无缝衔接啦。

install.packages("gganimate")

33.2.1 先来一张静态图 

library(tidyverse)
library(covdata) # remotes::install_github("kjhealy/covdata")
library(gganimate)
covdata::covnat %>%
  dplyr::filter(iso3 == "USA") %>%
  dplyr::filter(cu_cases > 0) %>%
  ggplot(aes(x = date, y = cases)) +
  geom_path() +
  labs(
    title = "美国新冠肺炎累积确诊病例",
    subtitle = "数据来源https://kjhealy.github.io/covdata/"
  )

让它动起来，我们只需要增加一行代码！

covdata::covnat %>%
  dplyr::filter(iso3 == "USA") %>%
  dplyr::filter(cu_cases > 0) %>%
  ggplot(aes(x = date, y = cases)) +
  geom_path() +
  labs(
    title = "美国新冠肺炎累积确诊病例 {frame_along}",
    subtitle = "数据来源https://kjhealy.github.io/covdata/"
  ) +
  transition_reveal(along = date)

33.2.2 相对复杂点的例子 

library(datasauRus)
ggplot(datasaurus_dozen) +
  aes(x, y, color = dataset) +
  geom_point()

用分面展示

ggplot(datasaurus_dozen) +
  aes(x, y, color = dataset) +
  geom_point() +
  facet_wrap(~dataset)

可以用动图展示

ggplot(datasaurus_dozen) +
  aes(x, y, color = dataset) +
  geom_point() +
  transition_states(dataset, 3, 1) + # <<
  labs(title = "Dataset: {closest_state}")

是不是很炫酷，下面我们就一个个讲解其中的函数。

33.3 The grammar of animation

使用gganimate做动画，只需要掌握以下五类函数：

  • transition_*(): 定义动画是根据哪个变量进行”动”，以及如何”动”
  • view_*(): 定义坐标轴随数据变化.
  • shadow_*(): 影子（旧数据的历史记忆）?定义点相继出现的方式.
  • enter_*()/exit_*(): 定义新数据出现和旧数据退去的方式.
  • ease_aes(): 美观定义，控制变化的节奏(如何让整个动画看起来更舒适).

下面通过案例依次讲解这些函数功能。

33.4 希望动画随哪个变量动起来

变量如何选择，这需要从变量类型和变量代表的信息来确定。

33.4.1 transition_states

  • transition_states(states = ), 这里的参数states往往带有分组信息，可以等价于静态图中的分面。
diamonds %>%
  ggplot(aes(carat, price)) +
  geom_point()
diamonds %>%
  ggplot(aes(carat, price)) +
  geom_point() +
  facet_wrap(vars(color))
diamonds %>%
  ggplot(aes(carat, price)) +
  geom_point() +
  transition_states(states = color, transition_length = 3, state_length = 1)

33.4.2 transition_time

  • transition_time(time = ), 这里的time一般认为是连续的值，相比于transition_states，没有了transtion_length这个选项，是因为transtion_length默认为time. 事实上，transition_time是transition_states的一种特例，但其实也有分组的要求
## Warning: package 'ggplot2' was built under R version 4.2.3
## Warning: package 'gganimate' was built under R version 4.2.2
## Warning: package 'gapminder' was built under R version 4.2.2
plot <- ggplot(gapminder, aes(gdpPercap, lifeExp, size = pop, color = country)) +
  geom_point(alpha = 0.7, show.legend = FALSE) +
  scale_colour_manual(values = gapminder::country_colors) +
  scale_size(range = c(2, 12)) +
  scale_x_log10() +
  facet_wrap(vars(continent)) +
  labs(title = "Year: {frame_time}", x = "GDP Per capita", y = "life expectancy") +
  transition_time(year) +
  ease_aes("linear")
plot
# save as a GIF
# animate(plot, fps = 30, width = 750, height = 750)
# anim_save("gapminder.gif")
p <- gapminder::gapminder %>%
  ggplot(aes(x = gdpPercap, y = lifeExp, size = pop, colour = country)) +
  geom_point(alpha = 0.7, show.legend = FALSE) +
  scale_size(range = c(2, 12)) +
  scale_x_log10() +
  labs(
    x = "GDP per capita",
    y = "life expectancy"
  )
p
anim <- p +
  transition_time(time = year) +
  labs(title = "year: {frame_time}")
anim

33.4.3 transition_reveal

  • transition_reveal(along = ), along 这个词可以看出，它是按照某个变量依次显示的意思，比如顺着x轴显示
ggplot(data = economics) +
  aes(x = date, y = unemploy) +
  geom_line()
ggplot(economics) +
  aes(x = date, y = unemploy) +
  geom_line() +
  transition_reveal(along = date) +
  labs(title = "now is {frame_along}")

33.4.4 transition_filter

  • transition_filter( 至少2个筛选条件，transition_length = , filter_length =), 动图将会在这些筛选条件对应的子图之间转换
diamonds %>%
  ggplot(aes(carat, price)) +
  geom_point() +
  transition_filter(
    transition_length = 3,
    filter_length = 1,

    cut == "Ideal",
    Deep = depth >= 60
  )

33.4.5 transition_layers 

mtcars %>%
  ggplot(aes(mpg, disp)) +
  geom_point() +
  geom_smooth(colour = "grey", se = FALSE) +
  geom_smooth(aes(colour = factor(gear))) +
  transition_layers(
    layer_length = 1, transition_length = 2,
    from_blank = FALSE, keep_layers = c(Inf, 0, 0)
  ) +
  enter_fade() +
  exit_fade()

33.5 希望坐标轴随数据动起来

动画过程中，绘图窗口怎么变化呢？

33.5.1 view_follow 

ggplot(iris, aes(Sepal.Length, Sepal.Width)) +
  geom_point() +
  labs(title = "{closest_state}") +
  transition_states(Species, transition_length = 4, state_length = 1) +
  view_follow()

33.6 希望动画有个记忆

  • shadow_wake(wake_length =, ) 旧数据消退时，制造点小小的尾迹的效果（wake除了叫醒，还有尾迹的意思，合起来就是记忆_尾迹）
  • shadow_trail(distance = 0.05) 旧数据消退时，制造面包屑一样的残留痕迹（记忆_零星残留）
  • shadow_mark(past = TRUE, future = FALSE) 将旧数据和新数据当作背景（记忆_标记）

33.6.1 shadow_wake() 

p +
  transition_time(time = year) +
  labs(title = "year: {frame_time}") +
  shadow_wake(wake_length = 0.1, alpha = FALSE)
ggplot(iris, aes(Petal.Length, Sepal.Length)) +
  geom_point(size = 2) +
  labs(title = "{closest_state}") +
  transition_states(Species, transition_length = 4, state_length = 1) +
  shadow_wake(wake_length = 0.1)

33.6.2 shadow_trail() 

p +
  transition_time(time = year) +
  labs(title = "year: {frame_time}") +
  shadow_trail(distance = 0.1)
ggplot(iris, aes(Petal.Length, Sepal.Length)) +
  geom_point(size = 2) +
  labs(title = "{closest_state}") +
  transition_states(Species, transition_length = 4, state_length = 1) +
  shadow_trail(distance = 0.1)

33.6.3 shadow_mark() 

p +
  transition_time(time = year) +
  labs(title = "year: {frame_time}") +
  shadow_mark(alpha = 0.3, size = 0.5)
ggplot(airquality, aes(Day, Temp)) +
  geom_line(color = "red", size = 1) +
  transition_time(Month) +
  shadow_mark(colour = "black", size = 0.75)

33.7 定义新数据出现和旧数据退去的方式

出现和退去的函数是成对的

33.7.1 enter/exit_fade()

透明度上的变化，我这里用柱状图展示，效果要明显一点。

tibble(
  x = month.name,
  y = sample.int(12)
) %>%
  ggplot(aes(x = x, y = y)) +
  geom_col() +
  theme(axis.text.x = element_text(angle = 45, hjust = 1, vjust = 1)) +
  transition_states(states = month.name)
tibble(
  x = month.name,
  y = sample.int(12)
) %>%
  ggplot(aes(x = x, y = y)) +
  geom_col() +
  theme(axis.text.x = element_text(angle = 45, hjust = 1, vjust = 1)) +
  transition_states(states = month.name) +
  shadow_mark(past = TRUE) +
  enter_fade()
p +
  transition_time(time = year) +
  labs(title = "year: {frame_time}") +
  enter_fade()

33.7.2 enter_grow()/exit_shrink()

大小上的变化

tibble(
  x = month.name,
  y = sample.int(12)
) %>%
  ggplot(aes(x = x, y = y)) +
  geom_col() +
  theme(axis.text.x = element_text(angle = 45, hjust = 1, vjust = 1)) +
  transition_states(states = month.name) +
  shadow_mark(past = TRUE) +
  enter_grow()
p +
  transition_time(time = year) +
  labs(title = "year: {frame_time}") +
  enter_grow() +
  enter_fade()

33.8 控制变化的节奏

控制数据点变化的快慢

p + ease_aes({aesthetic} = {ease})
p + ease_aes(x = "cubic")
knitr::include_graphics("images/ease.png")

Source: https://easings.net/

看下面的案例：

diamonds %>%
  ggplot(aes(carat, price)) +
  geom_point() +
  transition_states(color, transition_length = 3, state_length = 1) +
  ease_aes("cubic-in") # Change easing of all aesthetics
diamonds %>%
  ggplot(aes(carat, price)) +
  geom_point() +
  transition_states(color, transition_length = 3, state_length = 1) +
  ease_aes(x = "elastic-in") # Only change `x` (others remain “linear”)

33.9 标签

我们可能需要在标题中加入每张动画的信息，常用罗列如下

transition_states(states = ) +
  labs(title = "previous is {previous_state}, 
                current is {closest_state}, 
                next is {next_state}")

transition_layers() +
  labs(title = "previous is {previous_layers}, 
                current is {closest_layers}, 
                next is {next_layers}")

transition_time(time = ) +
  labs(title = "now is {frame_time}")


transition_reveal(along = ) +
  labs(title = "now is {frame_along}")

33.10 保存

33.10.1 Renderer options 

## # A tibble: 6 × 2
##   Function        Description                                          
##   <chr>           <chr>                                                
## 1 gifski_renderer Default, super fast gif renderer.                    
## 2 magick_renderer Somewhat slower gif renderer.                        
## 3 ffmpeg_renderer Uses ffmpeg to create a video from the animation.    
## 4 av_renderer     Uses the av package to create a video (using ffmpeg).
## 5 file_renderer   Dumps a list of image frames from the animation.     
## 6 sprite_renderer Creates a spritesheet from frames of the animation.

33.10.2 常用方法

一般用anim_save()保存为 gif 格式，方法类似ggsave()

animation_to_save <- diamonds %>%
  ggplot(aes(carat, price)) +
  geom_point() +
  transition_states(color, transition_length = 3, state_length = 1) +
  ease_aes("cubic-in")

anim_save("first_saved_animation.gif", animation = animation_to_save)

33.11 案例演示一

这是网上有段时间比较火的racing_bar图

ranked_by_date <- covdata::covnat %>%
  group_by(date) %>%
  arrange(date, desc(cu_cases)) %>%
  mutate(rank = 1:n()) %>%
  filter(rank <= 10) %>%
  ungroup()
ranked_by_date %>%
  filter(date >= "2020-05-01") %>%
  ggplot(
    aes(x = rank, y = cname, group = cname, fill = cname)
  ) +
  geom_tile(
    aes(
      y = cu_cases / 2,
      height = cu_cases,
      width = 0.9
    ),
    alpha = 0.8,
    show.legend = F
  ) +
  geom_text(aes(
    y = cu_cases,
    label = cname
  ),
  show.legend = FALSE
  ) +
  scale_x_reverse(
    breaks = c(1:10),
    label = c(1:10)
  ) +
  theme_minimal() +
  coord_flip(clip = "off", expand = FALSE) +
  labs(
    title = "日期: {closest_state}",
    x = "",
    caption = "Source: github/kjhealy/covdata"
  ) +
  transition_states(date,
    transition_length = 4,
    state_length = 1,
    wrap = TRUE
  ) +
  ease_aes("cubic-in-out")

33.12 案例演示二 

bats <- readr::read_csv("./demo_data/bats-subset.csv") %>%
  dplyr::mutate(id = factor(id))
bats %>%
  ggplot(aes(
    x = longitude,
    y = latitude,
    group = id,
    color = id
  )) +
  geom_point()

33.12.1 常规的方法 

bats %>%
  ggplot(aes(
    x = longitude,
    y = latitude,
    group = id,
    color = id
  )) +
  geom_point() +
  transition_time(time) +
  shadow_mark(past = TRUE)
  • geom_path()是按照数据点出现的先后顺序
  • geom_line()是按照数据点在x轴的顺序
bats %>%
  ggplot(aes(
    x = longitude,
    y = latitude,
    group = id,
    color = id
  )) +
  geom_path() +
  transition_time(time) +
  shadow_mark(past = TRUE)

33.12.2 炫酷点的 

bats %>%
  dplyr::mutate(
    image = "images/bat-cartoon.png"
  ) %>%
  ggplot(aes(
    x = longitude,
    y = latitude,
    group = id,
    color = id
  )) +
  geom_path() +
  ggimage::geom_image(aes(image = image), size = 0.1) +
  transition_reveal(time)

33.13 案例演示三

全球R-Ladies组织，会议活动的情况，我们在地图上用动图展示

rladies <- read_csv("./demo_data/rladies.csv")
rladies

这里需要一个地图，可以这样

ggplot() +
  ggplot2::borders("world", colour = "gray85", fill = "gray80") +
  ggthemes::theme_map()

当然，最好是这样

library(maps)
world <- map_data("world") 

world_map <- ggplot() +
  geom_polygon(data = world, 
               aes(x = long, y = lat, group = group),
               color = "white", fill = "gray80"
               ) +
  ggthemes::theme_map()

world_map

然后把点打上去

world_map +
  geom_point(
    data = rladies,
    aes(x = lon, y = lat, size = followers),
    colour = "purple", alpha = .5
  ) +
  scale_size_continuous(
    range = c(1, 8),
    breaks = c(250, 500, 750, 1000)
  ) +
  labs(size = "Followers")

用动图展示（这种方法常用在流行病传播的展示上）

world_map +
  geom_point(aes(x = lon, y = lat, size = followers),
    data = rladies,
    colour = "purple", alpha = .5
  ) +
  scale_size_continuous(
    range = c(1, 8),
    breaks = c(250, 500, 750, 1000)
  ) +
  transition_states(created_at) +
  shadow_mark(past = TRUE) +
  labs(title = "Day: {closest_state}")

33.14 课后作业

33.14.1 作业1

把下图弄成你喜欢的样子

library(gapminder)
theme_set(theme_bw())

ggplot(gapminder) +
  aes(
    x = gdpPercap, y = lifeExp,
    size = pop, colour = country
  ) +
  geom_point(show.legend = FALSE) +
  scale_x_log10() +
  scale_color_viridis_d() +
  scale_size(range = c(2, 12)) +
  labs(x = "GDP per capita", y = "Life expectancy") +
  transition_time(year) +
  labs(title = "Year: {frame_time}")

33.14.2 作业2

那请说说这以下三个的区别？

bats %>%
  dplyr::filter(id == 1) %>%
  ggplot(
    aes(
      x = longitude,
      y = latitude
    )
  ) +
  geom_point() +
  transition_reveal(time) # <<



bats %>%
  dplyr::filter(id == 1) %>%
  ggplot(
    aes(
      x = longitude,
      y = latitude
    )
  ) +
  geom_point() +
  transition_states(time) # <<



bats %>%
  dplyr::filter(id == 1) %>%
  ggplot(
    aes(
      x = longitude,
      y = latitude
    )
  ) +
  geom_point() +
  transition_time(time) # <<