highlight weekends using ggplot?

Question

I haven't found any history questions about this... I want to highlight the weekend performance for a ggplot graph, so that user could tell directly one the graph which part(may be a gray shade?) is weekend performance right away.

Here's a simple version of testing data:

test <- data.frame(DATE=seq(from = as.POSIXct("2014-07-16 01:00"), to = as.POSIXct("2014-07-30 00:00"), by = "hour"),count=floor(runif(336,1,100)))

Simple version of my graph is:

ggplot() + geom_line(aes(x=DATE,y=count),data=test) + labs(title="test")

So that result could be like something below...

enter image description here

See [geom_rect](http://docs.ggplot2.org/current/geom_rect.html)?. Should do what you're after. Also [this question](http://stackoverflow.com/questions/16504349/can-a-portion-of-the-background-in-ggplot-be-changed-to-a-different-color) may help. — Adam Kimberley, Aug 13 '14 at 15:49
@AdamKimberley Nice thanks! I can actually specify the date instead of the cell position. Think I need to come up a smarter way in longer time range but so far this works like a charm! :) — linp, Aug 13 '14 at 16:24

mynameisthis · Answer 1 · 2015-09-11T16:57:28.470

11

Using geom_area() is more concise, and will work for any date range.

    library(ggplot2)

    test <- data.frame(DATE=seq(from = as.POSIXct("2014-07-16 01:00"), 
                                to = as.POSIXct("2014-07-30 00:00"), 
                                by = "hour"),
                       count=floor(runif(336,1,100)))

    test$weekend <- weekdays(test$DATE) %in% c("Saturday", "Sunday")

    ggplot(data=test, aes(x=DATE, y=count)) +
      geom_area(aes(y=weekend*max(count)), fill="yellow") +
      geom_line() +
      labs(title="test")

edited Sep 11 '15 at 16:57

answered Sep 11 '15 at 02:35

mynameisthis

133
1
8

I couldn't get this to work for a faceted plot, but I found a useful clue here: https://stackoverflow.com/questions/51092214/faceting-based-on-geom-point-when-ggplot-empty-and-multiple-shaded-geom-rect-e – John Jan 23 '19 at 22:44

arvi1000 · Accepted Answer · 2015-02-17T17:29:58.883

5

Here's code that does both weekends in your data. You could generalize to any number of weekends by adding more geom_rect() calls (or calling a loop that does).

# your data
test <- data.frame(DATE=seq(from = as.POSIXct("2014-07-16 01:00"), to = as.POSIXct("2014-07-30 00:00"), by = "hour"),count=floor(runif(336,1,100)))
your_plot <- ggplot(test) + geom_line(aes(x=DATE,y=count)) + labs(title="test") 

# get all the start and end points
library(lubridate) # for hour function
sats <- which(hour(test$DATE)==0 & weekdays(test$DATE)=='Saturday')
suns <- which(hour(test$DATE)==23 & weekdays(test$DATE)=='Sunday')

# do your plot plus weekend highlights
your_plot +
  geom_rect(aes(xmin=DATE[sats[1]], xmax=DATE[suns[1]],
                  ymin=min(count), ymax=max(count)),
            fill='yellow', alpha=.005) +

  geom_rect(aes(xmin=DATE[sats[2]], xmax=DATE[suns[2]],
                ymin=min(count), ymax=max(count)),
            fill='yellow', alpha=.005)

Here's the output

edited Feb 17 '15 at 17:29

answered Aug 13 '14 at 16:22

arvi1000

9,393
2
42
52

Nice! I just followed Adam's suggestion specify the date, but this is exactly what I was looking for regarding to longer time range. Thanks! :) – linp Aug 13 '14 at 16:26
How about mentioning that the hours and weekdays functions are in the lubridate package? – mborowczak Feb 17 '15 at 17:23
Good point. Added. (Though it's just hour. weekdays is in base) – arvi1000 Feb 17 '15 at 17:30
You can also use `data.table::hour` – arvi1000 Feb 17 '15 at 17:38
If you're using a loop to add more geom_rect(), dont forget to use aes_string() instead of plain aes(). Otherwise only the last rectangle will be shown. – Daan Feb 19 '19 at 21:22

Romain · Answer 3 · 2020-03-22T09:45:04.670

Here is an approach using Tydiverse tools, and more precisely geom_tile. The advantage is that you have not to precise the x boundaries.

library(dplyr)
library(lubridate)
library(ggplot2)

test %>%
  # My preference :-)
  rename_all(tolower) %>%
  # Computing the weekends by converting in weekday starting on monday to ease the cut
  mutate(weekend = wday(date, week_start = getOption("lubridate.week.start", 1)) > 5) %>%
  ggplot() +
  geom_line(aes(x = date, y = count)) +
  # And here is the trick!
  scale_fill_manual(values = c("alpha", "yellow")) +
  geom_tile(aes(
    x = date,
    y = min(count),
    height = Inf,
    fill = weekend
  ), alpha = .4)

Note: I've written a more detailed post on this topic.

score 1 · Answer 4 · answered Aug 13 '14 at 16:19

As per my comment, hopefully this is what you are looking for:

test <- data.frame(DATE=seq(from = as.POSIXct("2014-07-16 01:00"), to = as.POSIXct("2014-07-30 00:00"), by = "hour"),count=floor(runif(336,1,100)))

ggplot() + geom_rect(aes(xmin = as.POSIXct("2014-07-26 00:00"), xmax = as.POSIXct("2014-07-28 00:00"), ymin = 0, ymax = 100), fill = "yellow")+
 geom_line(aes(x=DATE,y=count),data=test) + labs(title="test")

Giving pretty much exactly the image in the question. You can alter fill to change the colour and easily add further rectangles at different points if required.

Resulting image

score 1 · Answer 5 · edited Oct 15 '17 at 11:38

I found a nice solution with geom_bar() where you make a vector which is infinite on weekends, the remaining values are NA and will not be plotted. I used lubridate as I am more familiar with it. I also had to name the aesthetics in the first call to ggplot() so the axes inherit their proper names and y is not labelled as "weekend":

library(lubridate)
library(ggplot2)

test <- data.frame(DATE=seq(from = as.POSIXct("2014-07-16 01:00"), to = as.POSIXct("2014-07-30 00:00"), by = "hour"),count=floor(runif(336,1,100)))
your_plot <- ggplot(test) + geom_line(aes(x=DATE,y=count)) + labs(title="test") 
your_plot

test$wd <- wday(test$DATE, label = TRUE)
test$weekend[test$wd=="Sat" | test$wd=="Sun"] <- 1/0

highlight_plot <- ggplot(test, aes(x= DATE, y = weekend)) + 
  geom_bar(stat="Identity", aes(y=weekend), col="yellow", alpha=.4, width = 1) + 
geom_line(aes(x=DATE,y=count)) + labs(title="test") 

highlight_plot

This solution is pretty flexible and it's easy to adapt the method to highlight different vectors of dates.

highlight weekends using ggplot?

5 Answers5