Draw a lot of plots on the same canvas (clean way) - r

Question

I want to draw a number of similar plots with a loop.
What I do is:
plot(0, 0, type="l", col="white", xlim=range(1,N), ylim=range(0.5, 2.5)) # provide axes, frame, ...
for(col in colors)
{
X <- generate_X() # vector of random numbers
lines(1:N, X, type="l", col=col)
}
The problem is that random numbers sometimes go out of the range(0.5,2.5) and I want to lengthen ylim range. Atm I'm going to do it with min and max before plotting. But there must be much, much cleaner way which I poorly cant find anywhere.
I think I'm missing something basic about plotting, but I couldnt find the solution.
Thanks

Answer 1

I think there are two quick answers to the OP's question:
calculate the plot range before initializing the plot (implied by OP), or
use a "cleaner" plotting wrapper function.
Setup: First we need to define the variables and functions the OP implies and then generate some data to work with.
# Initialize our N number of X points and
# colors vector.
N <- 20
colors <- c("yellow", "red", "blue", "green")
# Create function 'generate_X' to perform
# as implied by the OP.
generate_X <- function(.N){
rnorm(n=.N, mean=0, sd=1)
}
# Generate the entire data frame
# using the 'matrix' function to shape
# the data quickly.
data <- data.frame(
id=1:N,
matrix(
generate_X(N*length(colors)),
ncol=length(colors)
)
)
The above code simply initializes the variables, function, and data needed for the OP's example.
Method 1: Calculate the plot range and initialize the plot. This is pretty easy using the 'range' function. In the data frame we created, there is an "id" column for our x values, so we use the range of 'data$id' for our x. Then, we find the range of all the data across every column EXCEPT the first column (data[,-1]) to find the overall y range. We initialize with the color white, since our background is also white. Otherwise, we would have a point in the lower-left and upper-right corners. I added x and y labels just for looks.
plot(
range(data$id),
range(data[,-1]),
col="white",
xlab="x",
ylab="y")
Next we just loop through and plot the lines.
for(i in 1:length(colors)){
lines(data$id, data[, i + 1], type="l", col=colors[i])
}
This is essentially the same thing the OP demonstrated, but it's adapted slightly to accept a data frame as input. It's far easier to reference columns using an integer counter (i in this case) rather than the list of colors.
Method 2: There are a lot of plot wrapper packages out there, and one of the most popular is the 'ggplot2' package, and for good reason. You can avoid a lot of the looping hassle with plots by feeding shaped data into a 'ggplot' function. The code here is much "cleaner" from a reading perspective.
# Load packages for shaping data and plotting.
library(reshape2)
library(ggplot2)
First, we need the 'reshape2' package, because we want to use "melted" data in our plot. This just makes the 'ggplot' code WAY cleaner. Then, we load up the 'ggplot2' package for the plotting.
For our plot, we initialize a plot without any instructions, so we can specify them in the geometry layer. If we were creating multiple layers from the same data, we would specify the options in the base plot layer, but for this, we are only creating a single geometry layer with lines. The + allows us to add plot layers.
Next, we choose a geometry layer ('geom_line' in this case) and specify the data as melt(data, id.vars="id"). This shapes our data for the 'ggplot' function to use with minimal code. We use the "id" column as the ID variable, since that contains our x values. The shaped data now looks more like this:
# id variable value
# 1 1 X1 -0.280035386
# 2 2 X1 -0.371020958
# 3 3 X1 -0.239889784
# 4 4 X1 0.450357442
# 5 5 X1 -0.801697283
# 6 6 X1 -0.453057841
# 7 7 X1 -0.451321958
# 8 8 X1 0.948124835
# 9 9 X1 2.724205279
# 10 10 X1 -0.725622824
# 11 11 X1 0.475545293
# 12 12 X1 0.533060822
# 13 13 X1 -1.928335572
# 14 14 X1 -0.466790259
# 15 15 X1 -1.606005895
# 16 16 X1 0.005678344
# 17 17 X1 -1.719827853
# 18 18 X1 0.601011314
# 19 19 X1 -2.056315661
# 20 20 X1 1.006169713
# 21 1 X2 -1.591227194
# ...
# 80 20 X4 -1.045224561
You don't need to get too hung up on the shaping. Just understand that "melted" data works better with the 'ggplot' functions. We specify our melted data as the data for our geometry layer, and then we use the 'aes' function to tell the geometry layer how to deal with our data. Our x values are in the "id" column, and our y values are in the "value" column. The next part is what removes the loops: we specify the color to be differentiated based on the "variable" column. In our melted data, the "variable" column contains the name of the column that the data originally came from, and using it to specify the color will tell 'ggplot' to automatically change the color for each new "variable" value.
ggplot() +
geom_line(
data=melt(data, id.vars="id"),
aes(
x=id,
y=value,
col=variable
),
lwd=1,
alpha=0.7)
I specified the line width ("lwd") and alpha values just to make the graph a little more readable.