Assigning geom_text from a different dataframe labels to a graph

Question

So I'm using Twitter APIs to gather info related to a certain topic, and one of the things I'm visualizing is the popularity of devices.

So far I have this: https://gyazo.com/441a9ab80b943f9e0c3a36131273844a

The above is generated by this code:

device_types_condensed <- (ggplot(manu_tweets3, aes(x= statusSource_clean , fill = isRetweet)) + geom_bar() 
                 + theme(panel.background=element_rect(fill='white'), 
                         axis.ticks.x=element_blank(),
                         axis.text.x=element_blank())
                         + theme(axis.ticks.x=element_blank(), axis.text.x = element_text(angle = 25),
                         axis.text=element_text(size=8)) 
                         + labs(x="", title = "Device Popularity for Tweet or Retweet Usage", y ="No. of Tweets on Device")
                         )
device_types_condensed

What I want to do is to add text above each bar that reflects the % of tweet activity that device is responsible for.

This means I am not changing the y-axis. The y-axis still reflects the count of tweet, and the number on top of the bar will be what reflects the percentage. So far I already have a table made with that value: https://i.gyazo.com/5f14d2c1352e8c9c2c5997678ceea3b4.png

What I can't figure out for the life of me is how to select the % labels in the table just above, and then apply them to the ggplot graph based on device type.

Sorry, don't have the rep to post images but I linked the URLs!

score 2 · Accepted Answer · answered Jul 20 '19 at 04:19

You're pretty close. I didn't have access to your exact data so I simplified your problem. You said you had some devices, each with a count of tweets associated with those devices, and that each device had a separate proportion associated with it. You also said these were in two different data.frames.

The most ggplot-ish way to handle this would be to join them together into a single data.frame because both data.frames share a common key: The device. This simplifies the ggplot2 code a touch. First, I'll work up a solution without combining, and then I will end by showing you how to combine your two data.frames together.

I generated data that looked similar to your data like this:

mydf <- data.frame(device = c("A", "B", "C"),
                   num_tweets = c(100, 200, 50))

prop_df <- data.frame(device = c("A", "B", "C"),
                      proportion = c(.29, .57, .14))

Without joining them together first, I think you can get what you want with code like this:

ggplot(mydf) +
  geom_col(aes(device,
               num_tweets)) +
  geom_text(data = prop_df,
            aes(device,
                max(mydf$num_tweets * 1.10),
                label = paste0(proportion * 100, "%"))) +
  scale_y_continuous(expand = expand_scale(mult = c(0, .1)))

Notice a few things:

I went with a geom_text call to get the percentages to display because I want ggplot2 to handle the x position for me (to match what already gets displayed when we call geom_col right above it) so the bars and percentages match up.
The geom_text call has as its first argument data = prop_df which tells geom_text to not use the the plot's default data.frame, mydf, and to use prop_df instead just for that layer.
In my aes call, I tell ggplot to map device to the x axis and then I hard-coded the y values to 110% of the maximum device count so they will display all at the same height, just above the bars.
ggplot2, by default, tries to shrink the plot area to match the data you've plotted and I wanted some more breathing room so I used expand_scale(mult = c(0, .1) to expand the plot in the y direction by 110%.

Is this similar to what you were looking for?

I then went ahead and simplified the ggplot call by joining the two data.frames together with dplyr::left_join prior:

library(dplyr)

mydf <- left_join(mydf, prop_df)

ggplot(mydf) +
  geom_col(aes(device,
               num_tweets)) +
  geom_text(aes(device,
                max(mydf$num_tweets * 1.10),
                label = paste0(proportion * 100, "%"))) +
  scale_y_continuous(expand = expand_scale(mult = c(0, .1)))

which is just a bit shorter and doesn't require you to override the data argument in geom_text.

What do you think?

Thank you so much, it seems combining the tables is the most elegant solution after all. I was tunnel-visioned away into trying another method to get it like that that had me pulling my hair out! — Yeahprettymuch, Jul 20 '19 at 04:40
You bet. It can take a few attempts before you can guess the least painful way to make your data work with ggplot. If my answer was sufficient, could you please mark it as a solution? — amoeba, Jul 20 '19 at 05:01
Using your suggestion I now got something like this, with the percentage amt just above the bar. Thanks a tonne!!! https://i.gyazo.com/c2162e5e39a2f5b293c5d4da534c8be5.png — Yeahprettymuch, Jul 20 '19 at 05:05

Assigning geom_text from a different dataframe labels to a graph

1 Answers1