Multiple plots using loops in R

Question

I'm still trying to get my head around using loops to plot in R. I would like to plot (any plot to visualise the data will do) columns z_1 against z_2 in the data frame below according to the different names in column x_1.

x_1 <- c("A1", "A1","A1", "B10", "B10", "B10","B10", "C100", "C100", "C100")


z_1 <- rnorm(10, 70) 

z_2 <- rnorm(10, 1.7)

A <- data.frame(x_1, z_1, z_2)

As such, I would like to end up with three different plots; one for category A1, one for B10 and another for C100. I can do this using three different codes but I would like to be able to use a loop or any other single code to execute all three plots on the same page. In reality, I have a large dataset (4,000 rows) and would like to plot a couple of IDs on a page (say 5 on a page).

I hope this makes sense. Thanks for your help.

Here's my attempt at plotting them individually:

for A1:

data_A1 <- A[which(A$x_1 == "A1"), ]
plot(data_A1$z_2, data_A1$z_1)

I also tried something like this but getting error messages

for ( i in A$x_1[[i]]){

plot(A[which(A$x_1==A$x_1[[i]]), ], aspect = 1)
}

Answer 1

A simple approach with loops would be

for (cat in unique(x_1)){
  d <- subset(A, x_1 == cat)
  plot(d$z_1, d$z_2)
}

unique(x_1) gets you all the unique values of x_1. Then, for each of these values get a corresponding subset and use this subset for plotting.

Answer 2

Just to understand why your original code didn't work:

Setting up data works fine

x_1 <- c("A1", "A1", "A1", "B10", "B10", "B10","B10", "C100", "C100", "C100")
z_1 <- rnorm(10, 70) 
z_2 <- rnorm(10, 1.7)
A <- data.frame(x_1, z_1, z_2)

The individual plot works fine, but as I said in a comment, the which is unnecessary

data_A1 <- A[which(A$x_1 == "A1"), ] # your way
plot(data_A1$z_2, data_A1$z_1)

data_A1 <- A[A$x_1 == "A1", ]    # deleting which() makes it cleaner
with(data_A1, plot(z_2, z_1))    # you can also use with() to save typing

Now the for loop. Let's review a simple for loop in R (pretty close to the example in ?"for"):

for (i in 1:5) {
   print(1:i)
}

Pretty straightforward, 1:5 is c(1, 2, 3, 4, 5), so first i is 1, then 2, etc. Your for loop has a problem in that first line:

for (i in A$x_1[[i]]) { ## already a problem

First i is A$x_1[[i]]? That won't work, i isn't defined yet. Also, A$x_1 is a vector, not a list, so you shouldn't be using [[ to subset it. But we don't want a subset yet, we want a vector of the values i should take. What we want in this case is for (i in c("A1", "B10", "C100")), but we also want to do it programmatically instead of typing out all the different possibilities. There's a couple common ways to get that:

unique(A$x_1) # as in Mark's solution
levels(A$x_1) # works because A$x_1 is a factor

We can put either of those expressions after the in. I changed your [[ to [ in the plot call. [[ is for lists only. I also took out the unnecessary which()

for (i in unique(A$x_1)) {   # this line is good
    plot(A[A$x_1==A$x_1[i], ], aspect = 1)  # still a problem
}

Let's remind ourselves what values i is taking: "A1", "B10", "C100". What's A$x_1 == A$x_1["A1"] going to give? Nothing useful.

for (i in unique(A$x_1)) {  
    plot(A[A$x_1 == i, ], aspect = 1)  # getting there
}

The above code plots something, and it's neat, but it's not what you want. There's a bunch of warnings, all of them telling us that aspect isn't a valid argument, so we'll delete it. Looking at the plot, you'll see that it's plotting 3 variables, because we haven't told it what to put on the x and y axes.

for (i in unique(A$x_1)) {   
    plot(A[A$x_1==i, "z_2"], A[A$x_1==i, "z_1"])  # z_2 on x, z_1 on y 
}   # Works!!!

Notice that this is almost identical to Mark's answer. You don't have to use i and j in for loops, he used cat. It's good practice to use a more descriptive name. Now let's fancy it up a little:

for (i in unique(A$x_1)) {   
    plot(A[A$x_1==i, "z_2"], A[A$x_1==i, "z_1"],
         xlim = range(A$z_2), ylim = range(A$z_1), # base the axes on full data range
         main = paste("Plot of", i))  # Give each a title
}

Next time: don't forget that you can run tiny pieces of code to see what they are. If you have a line like for (i in A$x_1[[i]]) that you're not sure if it's right, enter A$x_1[[i]] at the console, hopefully that will help you figure out that you haven't defined i, so you'll change it to

for (i in A$x_1)

then you run A$x_1 and realize it's length is 10. You want 3 graphs, not 10, so you need i to take 3 values, all of them different, etc.

Answer 3

Perhaps you don´t need a loop. Try using ggplots facet_grid(). Here is the documentation, full of examples.

library(ggplot2)
library(reshape2)

melted_a <- melt(A)


ggplot(melted_a, aes(variable, value)) +
  geom_jitter() +
  facet_grid(. ~ x_1)

ggplot(melted_a, aes(variable, value)) +
  geom_jitter() +
  facet_grid(variable ~ x_1)

Edit Perhaps this solves this problem. But if you need to do many plots that have a similar structure, you could make a function and use aes_string() instead of aes(). Note: I'm not an expert at writing functions, so probably someone could edit and improve it. (not tested)

ggplot_fun <- function(data, x, y, rowfacet, colfacet, ...){
  p <- ggplot(data, aes_string(x, y))
  p <- p + geom_jitter()
  p <- p + facet_grid(as.formula(sprintf("%s ~ %s", rowfacet, colfacet))
}

ggplot_fun(melted_a, variable, value, variable, x_1)

Idea taken from this question.