Specify scale_x_discrete for cut of continuous variabale

Question

I have a continuous time variable and I want to plot a heatmap of values. To do so using ggplot and geom_tile() I bin the time into time_bin and parse the bounds of the resulting intervals. My question is somewhat related to this one, which has no answer.

Here's a toy data.frame

grouped_df [36 × 5] (S3: grouped_df/tbl_df/tbl/data.frame)
 $ y       : chr [1:36] "1" "1" "1" "1" ...
 $ time_bin: Factor w/ 100 levels "(-0.03,0.3]",..: 1 2 3 4 1 2 3 4 1 2 ...
 $ value   : num [1:36] -0.4 -0.512 -0.608 -0.725 0.757 ...
 $ low     : num [1:36] -0.03 0.3 0.601 0.901 -0.03 0.3 0.601 0.901 -0.03 0.3 ...
 $ high    : num [1:36] 0.3 0.601 0.901 1.2 0.3 0.601 0.901 1.2 0.3 0.601 ...
 - attr(*, "groups")= tibble [9 × 2] (S3: tbl_df/tbl/data.frame)
  ..$ y    : chr [1:9] "1" "2" "3" "4" ...
  ..$ .rows: list<int> [1:9] 
  .. ..$ : int [1:4] 1 2 3 4
  .. ..$ : int [1:4] 5 6 7 8
  .. ..$ : int [1:4] 9 10 11 12
  .. ..$ : int [1:4] 13 14 15 16
  .. ..$ : int [1:4] 17 18 19 20
  .. ..$ : int [1:4] 21 22 23 24
  .. ..$ : int [1:4] 25 26 27 28
  .. ..$ : int [1:4] 29 30 31 32
  .. ..$ : int [1:4] 33 34 35 36
  .. ..@ ptype: int(0) 
  ..- attr(*, ".drop")= logi TRUE
> dput(toy)
structure(list(y = c("1", "1", "1", "1", "2", "2", "2", "2", 
"3", "3", "3", "3", "4", "4", "4", "4", "5", "5", "5", "5", "6", 
"6", "6", "6", "7", "7", "7", "7", "8", "8", "8", "8", "9", "9", 
"9", "9"), time_bin = structure(c(1L, 2L, 3L, 4L, 1L, 2L, 3L, 
4L, 1L, 2L, 3L, 4L, 1L, 2L, 3L, 4L, 1L, 2L, 3L, 4L, 1L, 2L, 3L, 
4L, 1L, 2L, 3L, 4L, 1L, 2L, 3L, 4L, 1L, 2L, 3L, 4L), .Label = c("(-0.03,0.3]", 
"(0.3,0.601]", "(0.601,0.901]", "(0.901,1.2]", "(1.2,1.5]", "(1.5,1.8]", 
"(1.8,2.1]", "(2.1,2.4]", "(2.4,2.7]", "(2.7,3]", "(3,3.3]", 
"(3.3,3.6]", "(3.6,3.9]", "(3.9,4.2]", "(4.2,4.5]", "(4.5,4.81]", 
"(4.81,5.11]", "(5.11,5.41]", "(5.41,5.71]", "(5.71,6.01]", "(6.01,6.31]", 
"(6.31,6.61]", "(6.61,6.91]", "(6.91,7.21]", "(7.21,7.51]", "(7.51,7.81]", 
"(7.81,8.11]", "(8.11,8.41]", "(8.41,8.71]", "(8.71,9.01]", "(9.01,9.31]", 
"(9.31,9.61]", "(9.61,9.91]", "(9.91,10.2]", "(10.2,10.5]", "(10.5,10.8]", 
"(10.8,11.1]", "(11.1,11.4]", "(11.4,11.7]", "(11.7,12]", "(12,12.3]", 
"(12.3,12.6]", "(12.6,12.9]", "(12.9,13.2]", "(13.2,13.5]", "(13.5,13.8]", 
"(13.8,14.1]", "(14.1,14.4]", "(14.4,14.7]", "(14.7,15]", "(15,15.3]", 
"(15.3,15.6]", "(15.6,15.9]", "(15.9,16.2]", "(16.2,16.5]", "(16.5,16.8]", 
"(16.8,17.1]", "(17.1,17.4]", "(17.4,17.7]", "(17.7,18]", "(18,18.3]", 
"(18.3,18.6]", "(18.6,18.9]", "(18.9,19.2]", "(19.2,19.5]", "(19.5,19.8]", 
"(19.8,20.1]", "(20.1,20.4]", "(20.4,20.7]", "(20.7,21]", "(21,21.3]", 
"(21.3,21.6]", "(21.6,21.9]", "(21.9,22.2]", "(22.2,22.5]", "(22.5,22.8]", 
"(22.8,23.1]", "(23.1,23.4]", "(23.4,23.7]", "(23.7,24]", "(24,24.3]", 
"(24.3,24.6]", "(24.6,24.9]", "(24.9,25.2]", "(25.2,25.5]", "(25.5,25.8]", 
"(25.8,26.1]", "(26.1,26.4]", "(26.4,26.7]", "(26.7,27]", "(27,27.3]", 
"(27.3,27.6]", "(27.6,27.9]", "(27.9,28.2]", "(28.2,28.5]", "(28.5,28.8]", 
"(28.8,29.1]", "(29.1,29.4]", "(29.4,29.7]", "(29.7,30.1]"), class = "factor"), 
    value = c(-0.400237814865178, -0.511748154576217, -0.608275784372335, 
    -0.7247244523613, 0.757330374272958, 0.776718248240655, 0.790724987842227, 
    0.998768292393676, 1.77979507741577, 1.52945992260953, 1.23688248753942, 
    0.908341101522855, -1.19238496743088, -1.26961869177189, 
    -1.32516082947541, -1.38508403291503, -0.629615898978496, 
    -0.620941659841269, -0.654461627603879, -0.751893145646679, 
    2.37041367819975, 2.51316431734167, 2.79123079147629, 3.1167959195256, 
    -0.519653805023003, -0.660372294607146, -0.972330338954077, 
    -1.31768889807071, 2.11290890332694, 2.21939390456188, 2.21971276073297, 
    2.15853062905871, 0.0602363209937846, 0.296034218681414, 
    0.372831656177899, 0.273945402834637), low = c(-0.03, 0.3, 
    0.601, 0.901, -0.03, 0.3, 0.601, 0.901, -0.03, 0.3, 0.601, 
    0.901, -0.03, 0.3, 0.601, 0.901, -0.03, 0.3, 0.601, 0.901, 
    -0.03, 0.3, 0.601, 0.901, -0.03, 0.3, 0.601, 0.901, -0.03, 
    0.3, 0.601, 0.901, -0.03, 0.3, 0.601, 0.901), high = c(0.3, 
    0.601, 0.901, 1.2, 0.3, 0.601, 0.901, 1.2, 0.3, 0.601, 0.901, 
    1.2, 0.3, 0.601, 0.901, 1.2, 0.3, 0.601, 0.901, 1.2, 0.3, 
    0.601, 0.901, 1.2, 0.3, 0.601, 0.901, 1.2, 0.3, 0.601, 0.901, 
    1.2, 0.3, 0.601, 0.901, 1.2)), class = c("grouped_df", "tbl_df", 
"tbl", "data.frame"), row.names = c(NA, -36L), groups = structure(list(
    y = c("1", "2", "3", "4", "5", "6", "7", "8", "9"), .rows = structure(list(
        1:4, 5:8, 9:12, 13:16, 17:20, 21:24, 25:28, 29:32, 33:36), ptype = integer(0), class = c("vctrs_list_of", 
    "vctrs_vctr", "list"))), row.names = c(NA, -9L), class = c("tbl_df", 
"tbl", "data.frame"), .drop = TRUE))

To plot it, I use

toy %>% 
  ggplot() +
  geom_tile(aes(x=time_bin, y=y, fill=value), color=NA)+
  geom_vline(aes(xintercept = unique(time_bin[length(time_bin)/2])))

Which produces

I would like the x axis to be labeled only using the first portion of the breaks (-0.03, 0.3, ...). I have been trying to use scale_x_discrete(breaks, labels) with different approaches but haven't made much progress. The x scale is long in the real data, so ideally it I would be able to do something like scale_x_discrete(breaks=scales::pretty_breaks(5)), but that's also not working.

I also tried using the low and high values that I parse from the cut, but that creates tile plots that contain vertical white lines everywhere.

Update

Using insight from one of the answers, I used factor(low) because my parsing works well, but the proposed readr::parse_number() does not. This gets the labels into the proper format.

The remaining portion of the question would be, how to show less factor levels using scale_x_discrete(breaks = ...) ?

For example, this seems to decimate the axis, a bit cumbersome but kinda works (from here)

scale_x_discrete(breaks = function(x){x[c(rep(FALSE, 9), TRUE)]})+

EDIT

For completeness....I am parsing the results from cut using

parse_cuts <- function(x){
  # This f(x) is made to be used as
  # mutate(map_df(time_bin, function(.x) parse_cuts(as.character(.x))))
  out <- sapply(strsplit(x, "\\(|,|]"), function(qq) as.numeric(qq[-1]))[,1]
  names(out) <- c("low", "high")
  return(unlist(out))
}

TarJae · Answer 1 · 2022-10-11T16:28:39.697

2

We could use parse_number from readr package it is in tidyverse. To apply here we first have to transform time_bin to character:

Update: transformed x to factor(x)

library(tidyverse)
toy %>% 
  mutate(x = factor(parse_number(as.character(time_bin)))) %>% 
  ggplot() +
  geom_tile(aes(x=x, y=y, fill=value), color=NA)+
  geom_vline(aes(xintercept = unique(x[length(x)/2])))
  xlab("time_bin")

edited Oct 11 '22 at 16:28

answered Oct 11 '22 at 15:29

TarJae

72,363
6
19
66

This I like this, but in my actual example, the `vline` doesn't show and I think the call to `factor` is messing up my level ordering in my real data. It might be a failure to parse? For example, the bin `(21,21.3]` gets parsed as `2121.3000` – Matias Andina Oct 11 '22 at 15:41
Please see my edits about `scale_x_discrete()` – Matias Andina Oct 11 '22 at 15:51
just add factor to x like : ` mutate(x = factor(parse_number(as.character(time_bin))))`. See my update: – TarJae Oct 11 '22 at 16:27

score 1 · Answer 2 · answered Oct 11 '22 at 15:56

A possible way to solve this, using insight from other answers. Not the prettiest but somewhat gets the job done. It might be a good idea to do a labeller function, like label_decimate() to produce the subset.

toy %>% 
  mutate(x = factor(low)) %>%
  ggplot() +
  geom_tile(aes(x=x, y=y, fill=value), color=NA)+
  geom_vline(aes(xintercept = unique(x[length(x)/2])))+
  scale_x_discrete(breaks = function(x){x[c(rep(FALSE, 9), TRUE)]})

Ottie · Answer 3 · 2022-10-11T16:16:16.830

A simple base-r solution is to gsub the left edge of the interval out and cbind this to the data argument of ggplot. A slightly different regex (^.*,([-.0-9]+).$) would allow you to extract the right edge.

ggplot(cbind(toy, "time" = gsub("^.([-.0-9]+),.*$", "\\1", toy$time_bin))) +
    geom_tile(aes(x=time, y=y, fill=value), color=NA)+
    geom_vline(aes(xintercept = unique(time[length(time)/2])))

However, this is misleading, as the x axis "looks" continuous, but is in fact a series of stacked factors which just happen to be in order and roughly contiguous. A better approach is to turn the x aes into a genuine numeric, which will also let you specify your scale breaks more easily.

ggplot(cbind(toy, "time" = as.numeric(gsub("^.([-.0-9]+),.*$", "\\1", toy$time_bin)))) +
 geom_tile(aes(x=time, y=y, fill=value), color=NA)+
 geom_vline(aes(xintercept = unique(time[length(time)/2])))

Note however that this will leave gaps where the tile sizes don't quite match (the distance between -0.030 and 0.300 is bigger than that between 0.300 and 0.601, and this is correctly shown here). If this is true in your real data, you can either fudge the width of the geom_tile, or use geom_rect and specify xmax and xmin from the left and right edges of the cut interval.

ggplot(cbind(toy, 
             "time_l" = as.numeric(gsub("^.([-.0-9]+),.*$", "\\1", toy$time_bin)),
             "time_r" = as.numeric(gsub("^.*,([-.0-9]+).$", "\\1", toy$time_bin)),
             "y_l" = as.numeric(toy$y) - 0.5, 
             "y_r" = as.numeric(toy$y) + 0.5
             )) +
  geom_rect(aes(xmin=time_l, xmax=time_r, ymin=y_l, ymax=y_r, fill=value), color=NA)+
  geom_vline(aes(xintercept = unique(time_l[length(time_l)/2])))

Specify scale_x_discrete for cut of continuous variabale

Update

EDIT

3 Answers3