要在 ggplot2 中绘制的平均 3D 数据

Question

Let's say I have 3 vectors:

x=round(runif(1000,1,65))
y=round(runif(1000,1,65))
z=runif(1000,0,1)

These are stored within a Dataframe df as columns. X and Y are integers, and I am searching for a solution specific to integers as well as an extended solution for doubles.

I can make a 2D histogram of this dataset, but I only get the count of x,y in each 2D bin.

ggplot(df,aes(x=x,y=y)) + geom_bin2d() + theme_bw()

I tried with geom_tile as well, but it actually overlays the heatmaps on top of each others. The behavior would be correct if I manually average the dataset beforehand. Yet I would like a solution that either does this elegantly or directly. I think of a graph where I see the average/median/user-defined as color in a 2D layout.

The solution uses preferably ggplot2

Answer 1

The argument fill in aes will allow you to define the colour, when linked to stat = 'identity' in the geom_bin2d call (which takes the z value to define the fill):

ggplot(df, aes(x, y, fill = z)) + geom_bin2d(stat = 'identity')

EDIT :

I see that you were asking it to combine the mean/median or some form of calculation for each of the 65x65 squares. Not automatically done, but perhaps with dplyr this would be a potential solution:

library(ggplot2)
library(dplyr)

df <- tibble(x=round(runif(1000,1,65)),
             y=round(runif(1000,1,65)),
             z=runif(1000,0,1))

df %>% 
  group_by(x, y) %>%               ## These two lines make a new value from z,
  summarise(fill = mean(z)) %>%       ## as a calculation from combos of x and y
  ggplot(aes(x, y, fill = fill)) + 
  geom_bin2d(stat = 'identity')

EDIT 2 :

A further question below about a) using a continuous variable and b) adjusting the number of bins:

bins <- 30

df %>% 
  mutate(x1 = as.numeric(cut(x, bins)),
         y1 = as.numeric(cut(y, bins))) %>% 
  group_by(x1, y1) %>%
  summarise(fill = mean(z)) %>%
  ggplot(aes(x1, y1, fill = fill)) + 
  geom_bin2d(stat = 'identity') +
  scale_x_continuous(breaks = c(1,bins), labels = c(1, max(df$x)))+
  scale_y_continuous(breaks = c(1,bins), labels = c(1, max(df$y)))

This produces a graph with a variable number of bins. The labels up the axes are hardest to reproduce here though, at the moment they're just set to label the top and bottom values. Take off the bottom two lines and you'll get it labelled by bin number (1-30) at least.

Answer 2

This will calculate the average of z for all pairs of x/y coordinates and plot on color scale:

df = data.frame(x, y, z)

library(dplyr)
library(ggplot2)

df %>% group_by(x, y) %>% summarize(mean_z = mean(z)) %>% 
   ggplot(aes(x = x, y = y, fill = mean_z)) + geom_bin2d(stat = "identity")