将geom_point更改为渐变填充的形状/多边形

Question

我正在探索特定索引如何变化。该指数是对数转换量表上第10个百分点与第90个百分点之间的比率的度量。为了探索索引，我创建了这个数据集：

dt = 
  data_frame(p10 = sample(seq(0.1,1.25,0.05), replace = TRUE, 10000),
             p90 = sample(seq(0.25,1.75,0.05), replace = TRUE, 10000)) %>% 
  filter(p90 > p10) %>%
  mutate(p10 = log(p10), p90 = log(p90)) %>%
  mutate(index = p10/p90) %>%
  filter(abs(index) < 10) %>%
  select(index, p10, p90)
dt

我绘制了数据集，以查看随着对数转换的百分位数变化，索引如何变化。

dt %>%
  ggplot(aes(x = p10, y = p90)) +
  geom_point(aes(colour = index)) +

  geom_abline(colour = "black", size = 0.75) +
  geom_hline(colour = "black", yintercept = 0, size = 0.75) +
  geom_vline(colour = "black", xintercept = 0, size = 0.75) +

  scale_colour_distiller(type = "div", palette = 1) + 

  coord_equal() +
  xlim(-2.5,0.75) +
  ylim(-2.5,0.75) +
  theme_bw()

此图很好地说明了该点，适合我个人使用，但是对于演示而言并不理想。我希望可以将点替换为带有渐变填充的多边形。但是，弄清楚如何做到这一点超出了我的能力范围。此外，我不确定是否确实可行。有人会为此介意将我指向正确的方向吗？非常感谢！

Answer 1

您可以签出interp函数来查看它是否满足您的需求：

library(akima)

# interpolate data
dt.interp <- interp(x = dt$p10, y = dt$p90, z = dt$index,
                    duplicate = "mean",
                    nx = 100, ny = 100) # set nx / ny based on how fine your want the polygons to be

# convert results back to a data frame
dt.interp <- data.frame(
  p10 = rep(dt.interp$x, times = length(dt.interp$y)),
  p90 = rep(dt.interp$y, each = length(dt.interp$x)),
  index = as.vector(dt.interp$z)
)

# plot results, replacing geom_point with geom_tile & color scale with fill scale
dt.interp %>%
  ggplot(aes(x = p10, y = p90)) +      
  geom_tile(aes(fill = index)) +      
  geom_abline(colour = "black", size = 0.75) +
  geom_hline(colour = "black", yintercept = 0, size = 0.75) +
  geom_vline(colour = "black", xintercept = 0, size = 0.75) +      
  scale_fill_distiller(type = "div", palette = 1) +       
  coord_equal() +
  xlim(-2.5,0.75) +
  ylim(-2.5,0.75) +
  theme_bw()

（灰色区域是NA值，因为那里没有可插值的点。如果要隐藏它们，可以在na.omit()上对df.interp运行+------------+----------+------------+--------+------------+--------+ | Closed | ClosedNo | Open | OpenNo | Plan | PlanNo | +------------+----------+------------+--------+------------+--------+ | NULL | 0 | NULL | 0 | 2018-10-22 | 3 | | NULL | 0 | NULL | 0 | 2018-10-23 | 1 | | NULL | 0 | NULL | 0 | 2018-10-24 | 1 | | NULL | 0 | NULL | 0 | 2018-10-25 | 1 | | NULL | 0 | 2018-10-25 | 1 | NULL | 0 | | NULL | 0 | 2018-10-26 | 1 | NULL | 0 | | NULL | 0 | 2018-10-27 | 2 | NULL | 0 | | 2018-10-22 | 3 | NULL | 0 | NULL | 0 | | 2018-10-23 | 1 | NULL | 0 | NULL | 0 | | 2018-10-25 | 1 | NULL | 0 | NULL | 0 | +------------+----------+------------+--------+------------+--------+ 或将它们分配给其他对象颜色。

Answer 2

这是一个将点转换为多边形并设置每个颜色的示例。我怀疑这正是您想要的，但这也许只是一个开始。

library(dplyr)
library(ggplot2)
library(sp)
library(rgeos)
dt = 
  data_frame(p10 = sample(seq(0.1,1.25,0.05), replace = TRUE, 10000),
             p90 = sample(seq(0.25,1.75,0.05), replace = TRUE, 10000)) %>% 
  filter(p90 > p10) %>%
  mutate(p10 = log(p10), p90 = log(p90)) %>%
  mutate(index = p10/p90) %>%
  filter(abs(index) < 10) %>%
  select(index, p10, p90)

dt.points = dt
coordinates(dt.points) = ~p10 + p90
dt.polygons = gBuffer(dt.points, width = 0.01, byid = T)

dt.polygons$colour = cut(dt$index, breaks = 10, labels = F)

plot(dt.polygons, col=dt.polygons$colour)