20

I want to create a heat map using ggplot however I want to order the y-axis by the number of observations. I order the dataframe by the column N and add the number of observations to the group name so that it appears in the axis label. When I plot the data it re-orders based on the group name. Is there a way to set factor levels based on the order they appear in the data frame?

Some data:

library(dplyr) library(tidyr) library(ggplot2) school <- c("School A", "SChool B", "School C", "School D", "School E", "School F") N <- c(25,28,12,22,30,25) var1 <- c(1,0,1,1,0,1) var2 <- c(0,0,0,1,0,1) var3 <- c(0,1,0,1,1,1) df <- tbl_df (data.frame (school, N, var1, var2, var3)) df <- arrange (df, N) %>% gather (variable, value, var1:var3) df$school <- paste0 (df$school, " (", df$N, ")") df <- select (df, school, variable, value) ggplot(df, aes(variable, school)) + geom_tile(aes(fill = value), colour = "white") + scale_fill_gradient(low = "white",high = "steelblue")

Ultimately I want the order of schools to be:

School C (12)

School D (22)

School A (25)

School F (25)

School B (28)

School E (30)

As I want to do this for multiple plots I want to find a way to do this automatically and not have to re-set factor levels each time.

Improve this question
0

3 Answers 3

Reset to default
17

One way around this is to change your ggplot call to 

ggplot(df, aes(variable, factor(school, levels = unique(school)))) + ...

To avoid typing this every time, you can create a function 

f <- function(x) factor(x, levels = unique(x))

and then call it by ggplot(df, aes(variable, f(school))) + ...

Note that this will place the first level of the factor at the bottom of the plot. If you want it at the top, you need to change f to function(x) factor(x, levels = rev(unique(x)))

Improve this answer
Sign up to request clarification or add additional context in comments.

2 Comments

How does factor(school, levels = unique(school)) change the result, compared to just factor(school)? I can see it does, but how does it link the order to the value N? And factor(df$school) == factor(df$school2, levels = unique(df$school2)) returns a vector TRUE of length 18 - how can that be so and yet have different results on the ggplot?
@James factor() does by default first sort the levels (alphabetically) -> see the documentation ?factor. If we use factor(school, levels = unique(school)), we force the levels to be in the order they appear in the vector, i.e. without first being sorted. You can have a look at the following example, where we start with an (alphabetically) unsorted vector: school_rev <- c( "School C", "School B", "School A") You'll notice the different order of the levels of: f1 <- factor(school_rev) and f2 <- factor(school_rev, unique(school_rev)).
15

Add the following forcats pipe to the code just before the call to ggplot().

library(forcats) df$school <- fct_inorder(df$school) %>% fct_rev()

fct_inorder() creates factor levels in data frame order and fct_rev() reverses them so the plot goes in the right direction.

Improve this answer

1 Comment

This worked well for my use case of ordering a character based on another field with the "duplicate" levels that couldn't be applied to a factor. I ended up sorting my data using arrange followed by this code. Now I can get back to this complicated plot!
2

One way would be to make the school column and ordered factor:

df$school <- reorder(df$school, rep(6:1, length.out=length(k)), order=TRUE)

enter image description here

Improve this answer

Comments

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.