Heat map per column with ggplot2

Question

I'm using this R script:

tableau <- read.table(
  text = 
    "Net    B   C   D   E.(e)   F.(f)
a   1.88    0.15    0.60    10.00   90.00
b   2.05    0.23    0.51    55.00   80.00
c   2.09    0.29    0.40    58.00   88.00
d   2.07    0.52    0.36    80.00   84.00
e   2.13    0.30    0.27    7.00    90.00",
  header = TRUE)

library(plyr)
library(reshape)
library(ggplot2)
library(scales)
tableau.m <- melt(tableau)
tableau.m <- ddply(tableau.m, .(variable), transform, rescale = rescale(value))

(p <- ggplot(tableau.m, aes(variable, Net)) + 
    geom_tile(aes(fill = rescale), colour = "white") + 
    scale_fill_gradient(low = "white", high = "steelblue"))

base_size <- 9
p + theme_grey(base_size = base_size) + 
  labs(x = "", y = "") + scale_x_discrete(expand = c(0, 0)) + 
  scale_y_discrete(expand = c(0, 0)) + 
  theme(legend.position = "none", axis.ticks = element_blank(), 
        axis.text.x = element_text(size = base_size * 0.8, angle = 0, 
                                   hjust = 0, colour = "grey50"))

tableau.s <- ddply(tableau.m, .(variable), transform, rescale = scale(value))

last_plot() %+% tableau.s

And I obtain this plot:

Where darker blue means higher values and white means lower values.

How, if possible, could I change this code so that:

1. the values from the table are displayed in each corresponding cell of the matrix plot?
2. the range of the heat map isn't calculated on the whole matrix, but rather for each column. So that, for each category: B, C, D, E(e), and F(f), white means the lower value for this column, and darker blue means the higher value of the column?

Thanks!

Answer 1

To add value as a text label to each cell, you can use geom_text:

p <- ggplot(tableau.m, aes(variable, Net)) + 
      geom_tile(aes(fill = rescale), colour = "white") + 
      scale_fill_gradient(low = "white", high = "steelblue") +
      geom_text(aes(label=value))

# Add the theme formatting
base_size <- 9
p + theme_grey(base_size = base_size) + 
  labs(x = "", y = "") + scale_x_discrete(expand = c(0, 0)) + 
  scale_y_discrete(expand = c(0, 0)) + 
  theme(legend.position = "none", axis.ticks = element_blank(), 
        axis.text.x = element_text(size = base_size * 0.8, 
                                   angle = 0, hjust = 0, colour = "grey50"))

For your second question, your current code already takes care of that. The variable rescale scales each column separately, because you've performed the operation grouped by variable. Since rescale is the fill variable, each column's values are rescaled from zero to one for the purposes of setting color values. You don't need the tableau.s ... last.plot... code.

Here's what the plot looks like after running the code above. Note that in each column, the lowest value is white and the highest value is steel blue. (You might want to change the border color from "white" to, say, "gray90", so that there will be a border between adjacent white squares):

Answer 2

Similar idea using tidyr and dplyr to reshape the data to long format and ggvis to plot the heatmap:

library(dplyr)
library(ggvis)
library(tidyr)

tableau %>% 
  gather(variable, value, -Net) %>%
  group_by(variable) %>%
  mutate(scale = percent_rank(value)) %>%
  mutate_each(funs(factor(.)), -value, -scale) %>%
  ggvis(~variable, ~Net, fill=~scale) %>%
  layer_rects(width = band(), height = band(), stroke := NA) %>%
  layer_text(
    x = prop("x", ~variable, scale = "xcenter"),
    y = prop("y", ~Net, scale = "ycenter", ),
    text:=~value, fontSize := 14, fontWeight := "bold", fill:="black", 
    baseline:="middle", align:="center") %>%
  scale_nominal("x", padding = 0, points = FALSE) %>%
  scale_nominal("y", padding = 0, points = FALSE) %>% 
  scale_nominal("x", name = "xcenter", padding = 1, points = TRUE) %>%
  scale_nominal("y", name = "ycenter", padding = 1, points = TRUE) %>%
  scale_numeric("fill", range = c("white", "steelblue")) %>%
  add_axis("x", properties = axis_props(grid = list(stroke = NA))) %>%
  add_axis("y", properties = axis_props(grid = list(stroke = NA))) %>%
  hide_legend("fill")

Which gives: