Boxplot for large matrix crashes computer

I have a very large matrix A of size 3831360x28. I would like to make a boxplot for all 28 columns side by side to compare, but whenever I run the command

boxplot(A)

it uses all of my computer's memory and I need to kill the power to regain control. Is there a way around this? By doing some experimenting, my computer crashes somewhere between 1 and 10 million data points being put into this command.

Answer 1

[the cyclist]

I'm able to run that command without a crash. (R2015a, on a fairly new Macbook Pro.)

One possible solution would be for you to sample your data (say, 1 out of every 10 points) before plotting.

[Max Gawryla]

I'm also running R2015a, but I'm on a fairly old Dell laptop. What exactly do you mean by "sample my data"?

EDIT: I understand how to sample data, in this particular situation, the devil in the details and sampling will really blur those fringe events.

[the cyclist]

I guess it depends on exactly what you are trying to accomplish with the box plot. I assumed you were just getting a general feel for the different distributions, or convey those to some audience. If you choose 1 out of 10 points at random to plot, I would expect the "look and feel" of the plots will be nearly identical, good enough for preliminary exploratory work.

What are you trying to actually do, if "blurring fringe events" matters? I would think some analytical technique would be needed, not just a plot.