How to set Bin Width With geom_bar stat="identity" in a time Series plot?

Question

I would like to plot a time series using bar charts and have the Bin Width set to 0.9. I cannot seem to be able to do that however. I have searched around but could not find anything helpful so far. Is this a limitation if the stat="identity ?

Here is a sample data and graph. Cheers !

time <- c('2015-06-08 00:59:00','2015-06-08 02:48:00','2015-06-08 06:43:00','2015-06-08 08:59:00','2015-06-08 10:59:00','2015-06-08 12:59:00','2015-06-08 14:58:00','2015-06-08 16:58:00','2015-06-08 18:59:00','2015-06-08 20:59:00','2015-06-08 22:57:00','2015-06-09 00:59:00','2015-06-09 01:57:00','2015-06-09 03:22:00','2015-06-09 06:14:00','2015-06-09 08:59:00','2015-06-09 10:59:00','2015-06-09 12:59:00','2015-06-09 14:59:00','2015-06-09 16:59:00','2015-06-09 18:59:00','2015-06-09 20:59:00','2015-06-09 22:58:00','2015-06-10 00:57:00','2015-06-10 02:34:00','2015-06-10 04:45:00','2015-06-10 06:24:00','2015-06-10 08:59:00','2015-06-10 10:59:00','2015-06-10 12:59:00','2015-06-10 14:59:00','2015-06-10 16:59:00','2015-06-10 18:59:00','2015-06-10 20:58:00','2015-06-10 22:52:00','2015-06-11 00:59:00','2015-06-11 02:59:00','2015-06-11 04:59:00','2015-06-11 06:59:00','2015-06-11 08:59:00','2015-06-11 10:59:00','2015-06-11 12:59:00','2015-06-11 14:59:00','2015-06-11 16:58:00','2015-06-11 18:58:00','2015-06-11 20:56:00','2015-06-11 21:49:00','2015-06-12 00:59:00','2015-06-12 02:59:00','2015-06-12 04:20:00','2015-06-12 08:55:00','2015-06-12 10:55:00','2015-06-12 12:59:00','2015-06-12 14:59:00','2015-06-12 16:59:00','2015-06-12 18:59:00','2015-06-12 20:55:00','2015-06-12 22:50:00','2015-06-13 00:16:00','2015-06-13 12:59:00','2015-06-13 14:35:00','2015-06-13 16:56:00','2015-06-13 18:59:00','2015-06-13 20:59:00','2015-06-13 22:44:00','2015-06-13 23:19:00','2015-06-14 08:53:00','2015-06-14 10:14:00','2015-06-14 12:59:00','2015-06-14 14:59:00','2015-06-14 16:56:00','2015-06-14 18:58:00','2015-06-14 20:57:00','2015-06-14 22:31:00','2015-06-14 23:59:00')
count <- c(59,63,9,13,91,80,97,210,174,172,167,74,43,18,18,29,136,157,126,170,188,135,207,216,163,163,126,111,172,213,209,265,203,205,195,201,171,157,153,176,187,252,227,223,171,162,146,161,136,124,155,239,233,157,158,125,138,45,45,1,2,6,6,46,48,4,1,1,12,56,65,122,81,110,42)
level <- c('low','low','low','low','low','low','low','high','normal','normal','normal','low','low','low','low','low','low','normal','low','normal','normal','low','high','high','normal','normal','low','low','normal','high','high','high','high','high','normal','high','normal','normal','normal','normal','normal','high','high','high','normal','normal','low','normal','low','low','normal','high','high','normal','normal','low','low','low','low','low','low','low','low','low','low','low','low','low','low','low','low','low','low','low','low')
DF = data.frame(time, count, level) 
DF$time = as.POSIXct(DF$time)

ggplot(DF, aes(x=time, y=count, fill=level), width=0.9) + 
  geom_bar(stat="identity") + 
  scale_x_datetime(labels = date_format("%D"), breaks = date_breaks("day")) + 
  xlab("myXlabel") +
  ylab("myYlabel") +
  ggtitle("myTitle") 

Davidadavidde

When you say you want to set the bin width, do you mean you want to adjust the width (horizontal dimension) of the bars? – Adrian

@tegancp, Yes that's what I want indeed ! – Davidadavidde

Answer 1

Found it ! Actually, the width is supported, though the scale is in seconds since I'm plotting a time series where the X axis is formatted as a POSIX date. Therefore, a width=0.9 means the bin width is 0.9 seconds. Since my bins are 2hrs eachs then a width of "1" is actually 7200. So here is the code that works.

ggplot(DF, aes(x=time, y=count, width=6000, fill=level)) + 
  geom_bar(stat="identity", position="identity", color="grey") + 
  scale_x_datetime(labels = date_format("%D"), breaks = date_breaks("day")) + 
  xlab("myXlabel") +
  ylab("myYlabel") +
  ggtitle("myTitle") 

Results as below. There are some averlaps in the bars, I just need to aligh my data, say to the next hour.

Davidadavidde

Answer 2

If what you are trying to achieve is widening the bars in the plot, ggplot doesn't seem to support that for geom_bar. However, it is pretty straightforward to implement a barplot using geom_rect.

Since many of the datapoints seem to be spaced roughly two hours apart, I am assuming here that the 0.9 width you want to achieve is 0.9 hours to either side of the given time (so basically smushing out most of the space between the bars.

If that's what you want, the following code should work:

library(lubridate)

ggplot(DF, aes(xmin=time-minutes(54), xmax=time+minutes(54), ymin=0, ymax=count,
           fill=level)) +
geom_rect(color="#666666")

Adrian

That's weird geom_bar does not support the width option, I'm sure there is a way ... – Davidadavidde

Answer 3

I am also trying to wrap my head around R. I have worked on a solution and found a solution that also provided me with a warning pointing at the problem - overlapping time x intervals. The error disappears at width = 2000. By supplementing with position = "dodge": "places overlapping objects directly beside one another" - https://r4ds.had.co.nz/data-visualisation.html - you can achieve a reasonable plot.

# Original file
ggplot(DF, aes(x=time, y=count, fill=level, width=2000), position = "dodge") + 
    geom_bar(stat="identity") + 
    scale_x_datetime(labels = date_format("%D"), breaks = date_breaks("day")) + 
    xlab("myXlabel") +
    ylab("myYlabel") +
    ggtitle("myTitle")

PREVIOUS VERSION NOT SO GOOD Here is another solution:

ggplot(DF, aes(x=time, y=count, colour = level)) +
    geom_bar(stat="identity") +
    scale_x_datetime(labels = date_format("%D"), breaks = date_breaks(width = "day")) +
    xlab("myXlabel") +
    ylab("myYlabel") +
    ggtitle("myTitle")

colour=level gives wider columns

Taskmaster

can you guess the reason for it? (tip: try color = "black", and you will see. ) Ergo: This approach is not a good idea. – Tartary

Error in FUN(X[[i]], ...) : object 'black' not found - Thanks learning a little every day. Still find I peculiar that the columns behave like they do. – Taskmaster

ok - you have not used the quotes as I have suggested above. Also try using color = "black" outside of the aes, like so: geom_bar(..., aes(...), color = "black"). – Tartary

What color does - it draws a border around the columns, in the respective color. It only seems wider, because you're mapping the same aesthetic to both fill and color, thus you don't see that there is an extra line around the bars. – Tartary

@Tartary Thanks again, I have been messing around with the problem and have ended up with splitting the POSIX column giving the possibility to use "date" and "hour" more flexible and being able to make many different and informative plots. I agree with you that my solution above is wrong - perhaps you could downvote it? – Taskmaster

Thanks again @Tartary and I have edited my proposal with another solution – Taskmaster