R | combine plots that use par(mfrow = ...) internally

Question

Take plot.acf as an example. Both acf and pacf call this function internally. How can i plot them side by side?

Example:

TS <- ts.union(mdeaths, fdeaths)
acf(TS)
pacf(TS)

I tried to use par(mfrow = c(2,4)) and layout to combine them, but stats:::plot.acf overwrites this. The expected output would be:

Answer 1

A different approach than my other answer: Plot the ACF using ggplot2.

ggacf <- function(x, ci=0.95, type="correlation", xlab="Lag", ylab=NULL,
                  ylim=NULL, main=NULL, ci.col="blue", lag.max=NULL) {

    x <- as.data.frame(x)

    x.acf <- acf(x, plot=F, lag.max=lag.max, type=type)

    ci.line <- qnorm((1 - ci) / 2) / sqrt(x.acf$n.used)

    d.acf <- data.frame(lag=x.acf$lag, acf=x.acf$acf)

    g <- ggplot(d.acf, aes(x=lag, y=acf)) +
        geom_hline(yintercept=0) +
        geom_segment(aes(xend=lag, yend=0)) +
        geom_hline(yintercept=ci.line, color=ci.col, linetype="dashed") +
        geom_hline(yintercept=-ci.line, color=ci.col, linetype="dashed") +
        theme_bw() +
        xlab("Lag") +
        ggtitle(ifelse(is.null(main), "", main)) +
        if (is.null(ylab))
            ylab(ifelse(type=="partial", "PACF", "ACF"))
        else
            ylab(ylab)

    g
}

This seeks to create a similar interface to plot.acf(). Then you can use all of the great features available to ggplot2 plots from the gridExtra package.

library(ggplot2)
library(gridExtra)

grid.arrange(ggacf(lh), ggacf(lh, type="partial"), ncol=2)

Then you get this:

Unfortunately grid.arrange() doesn't work with base graphics, hence the ggplot2 suggestion.