Question

I am trying to plot two histograms in one plot, but the way these two groups are distributed makes the histogram a little hard to interpret. My histogram now looks like this:

This is my code:

hist(GROUP1, col=rgb(0,0,1,1/2), breaks=100, freq=FALSE,xlab="X",main="")  # first histogram
hist(GROUP1, col=rgb(1,0,0,1/2), breaks=100, freq=FALSE , add=T)  # second
legend(0.025,600,legend=c("group 1","group 2"),col=c(rgb(1,0,0,1/2),rgb(0,0,1,1/2)),pch=20,bty="n",cex=1.5)

Is it possible to plot this histograms, with the bars of the two groups right next to each other, instead of them overlapping? I realize that that might add some confusion, since the X-axis represents a continuous variable… Other suggestions of how to make this plot in more clear are of course also welcome!

Answer 1

You could misuse barplot for it:

multipleHist <- function(l, col=rainbow(length(l))) {
    ## create hist for each list element
    l <- lapply(l, hist, plot=FALSE);

    ## get mids
    mids <- unique(unlist(lapply(l, function(x)x$mids)))

    ## get densities
    densities <- lapply(l, function(x)x$density[match(x=mids, table=x$mids, nomatch=NA)]);

    ## create names
    names <- unique(unlist(lapply(l, function(x)x$breaks)))

    a <- head(names, -1)
    b <- names[-1]
    names <- paste("(", a, ", ", b, "]", sep="");

    ## create barplot list
    h <- do.call(rbind, densities);

    ## set names
    colnames(h) <- names;

    ## draw barplot
    barplot(h, beside=TRUE, col=col);

    invisible(l);
}

Example:

x <- lapply(c(1, 1.1, 4), rnorm, n=1000)
multipleHist(x)

EDIT:
Here is an example to draw a x-axis like the OP suggested. IMHO this is very misleading (because bins of a barplot are not continuous values) and should not be used.

multipleHist <- function(l, col=rainbow(length(l))) {
    ## create hist for each list element
    l <- lapply(l, hist, plot=FALSE);

    ## get mids
    mids <- unique(unlist(lapply(l, function(x)x$mids)))

    ## get densities
    densities <- lapply(l, function(x)x$density[match(x=mids, table=x$mids, nomatch=NA)]);

    ## create names
    breaks <- unique(unlist(lapply(l, function(x)x$breaks)))

    a <- head(breaks, -1)
    b <- breaks[-1]
    names <- paste("(", a, ", ", b, "]", sep="");

    ## create barplot list
    h <- do.call(rbind, densities);

    ## set names
    colnames(h) <- names;

    ## draw barplot
    barplot(h, beside=TRUE, col=col, xaxt="n");

    ## draw x-axis
    at <- axTicks(side=1, axp=c(par("xaxp")[1:2], length(breaks)-1))
    labels <- seq(min(breaks), max(breaks), length.out=1+par("xaxp")[3])
    labels <- round(labels, digits=1)
    axis(side=1, at=at, labels=breaks)

    invisible(l);
}

Please find the complete source code on github.