Boxplots for groups?

Question

I have a dataset (test) as given below:

Type    Met1    Met2    Met3    Met4
TypeA   65  43  97  77
TypeA   46  25  76  77
TypeA   44  23  55  46
TypeA   46  44  55  77
TypeA   33  22  55  54
TypeB   66  8   66  47
TypeB   55  76  66  65
TypeB   55  77  88  46
TypeB   36  67  55  44
TypeB   67  55  76  65

I have checked a lot of links on box plots, but I still have not succeeded for the type of box plot I want. I wish to have a boxplot with my X-axis having type A (yellow, orange) for all the Mets (Met1, Met2, Met3, Met4). In essence, I want something like the following (taken from here):

I am trying somethings like,

boxplot(formula = len ~ Type , data = test, subset == "TypeA")
boxplot(formula = len ~ Type , data = test, subset == "TypeA", add=TRUE)
Legend(legend = c( "typeA", "typeB" ), fill = c( "yellow", "orange" ) )

But I am not able to work it out with any of it. Can anyone help me know how do I make such box plots on my test data in the corrected way?

Answer 1

A solution with ggplot2.

First, transform your data frame test to the long format using melt:

library(reshape2)
test.m <- melt(test)

Plot the data:

library(ggplot2)
ggplot(test.m, aes(x = variable, y = value, fill = Type)) +
  geom_boxplot() +
  scale_fill_manual(values = c("yellow", "orange"))

Answer 2

As others have said, first you need to melt your data.

df <- read.table(text="Type    Met1    Met2    Met3    Met4
TypeA   65  43  97  77
TypeA   46  25  76  77
TypeA   44  23  55  46
TypeA   46  44  55  77
TypeA   33  22  55  54
TypeB   66  8   66  47
TypeB   55  76  66  65
TypeB   55  77  88  46
TypeB   36  67  55  44
TypeB   67  55  76  65",header=TRUE)

library(reshape2)
df2 <- melt(df)

boxplot(
  formula = value ~ variable,
  data    = df2,
  boxwex  = 0.25,
  at      = 1:4 - 0.2,
  subset  = Type == "TypeA",
  col     = "yellow",
  main    = "blah",
  xlab    = "x",
  ylab    = "y",
  ylim    = c( 0, ceiling( max( df2$value ) ) + 1 ),
  yaxs    = "i" )


boxplot(
  formula = value ~ variable,
  data    = df2,
  boxwex  = 0.25,
  at      = 1:4 + 0.2,
  subset  = Type == "TypeB",
  col     = "orange",
  add     = TRUE )

Answer 3

like this,

test <- structure(list(Type = structure(c(1L, 1L, 1L, 1L, 1L, 2L, 2L, 
    2L, 2L, 2L), .Label = c("TypeA", "TypeB"), class = "factor"), 
    Met1 = c(65L, 46L, 44L, 46L, 33L, 66L, 55L, 55L, 36L, 67L
    ), Met2 = c(43L, 25L, 23L, 44L, 22L, 8L, 76L, 77L, 67L, 55L
    ), Met3 = c(97L, 76L, 55L, 55L, 55L, 66L, 66L, 88L, 55L, 
    76L), Met4 = c(77L, 77L, 46L, 77L, 54L, 47L, 65L, 46L, 44L, 
    65L)), .Names = c("Type", "Met1", "Met2", "Met3", "Met4"), 
    class = "data.frame", row.names = c(NA, -10L))


# install.packages("ggplot2", dependencies = TRUE)
require(ggplot2)
require(reshape2)
df <- melt(test)
p <- ggplot(df, aes(factor(variable), value)) + geom_boxplot(aes(fill = Type))
p

You take a look at the geom_boxplot manual page.

Answer 4

You can use the reshape function to make a long column with all the data from Met1, .., Met4 columns. It also creates a column (temp$time) to identify from which column the data came from that you can use it to stratify your box plot (temp$Type*temp$time).

df <- read.table(text=
"Type Met1 Met2 Met3 Met4
TypeA   65  43  97  77
TypeA   46  25  76  77
TypeA   44  23  55  46
TypeA   46  44  55  77
TypeA   33  22  55  54
TypeB   66  8   66  47
TypeB   55  76  66  65
TypeB   55  77  88  46
TypeB   36  67  55  44
TypeB   67  55  76  65",header=TRUE)

temp <- reshape(df, direction='long', varying = 2:5, sep='')

boxplot(temp$Met ~ temp$Type*temp$time, col=c("yellow", "orange"))