将多个glm图导出为PNG？

Question

SO,

I'm trying to export plots of my linear model. When I do it as a PDF, the PDF has four pages and four different charts. When I export as a PNG, I only get the first chart. How do I export so that I get all four graphs as separate PNG files?

What worked with the PDF:

lrfitX11SUMS <- glm(X11SUMS ~ POLITA + CIRCULATION + COMPAPER + TrafficRankUS, data=NewspTest)

    summary(lrfitOTONE)
    pdf("/Users/william/Desktop/output/lmfitOTONE1.pdf")
    plot(lrfitOTONE)
    dev.off()

What DIDN'T WORK with PNG (and spent two hours digging around on the internet and in the plot documentation to no avail):

lrfitX11SUMS <- glm(X11SUMS ~ POLITA + CIRCULATION + COMPAPER + TrafficRankUS, data=NewspTest)

summary(lrfitOTONE)
png("/Users/william/Desktop/output/lmfitOTONE1.png", width=720, height=720, pointsize=16)
png("/Users/william/Desktop/output/lmfitOTONE2.png", width=720, height=720, pointsize=16)
png("/Users/william/Desktop/output/lmfitOTONE3.png", width=720, height=720, pointsize=16)
png("/Users/william/Desktop/output/lmfitOTONE4.png", width=720, height=720, pointsize=16)
plot(lrfitOTONE)
dev.off()

How do I get my images?

Thanks,

-Wm

Answer 1

A PDF allows multipage documents. A PNG image is fundamentally incompatible with this idea. Reading ?png and appreciating the need to look at the filename argument would have directed you to ?postscript for the details.

You want something like:

png("/Users/william/Desktop/output/lmfitOTONE%1d.png", width=720, 
    height=720, pointsize=16)
plot(lrfitOTONE)
dev.off()

where the %1d in the filename is a wildcard that expands to a 1 digit numeric value such that you get four figures with the names you wanted. Your 4 calls to png() set up four separate devices, only the latter of which was used and subsequently closed, the other three remained open.

Answer 2

Something like this:

setwd("/Users/william/Desktop/output/")
tmpf <- function(i) {
   png(paste("lmfitOTONE",i,".png",sep=""), width=720, height=720, pointsize=16)
}
wplot <- c(1,2,3,5) ## see ?plot.lm for definition of 'which'
for (i in seq_along(wplot)) {
   tmpf(i); plot(lrfitOTONE, which=wplot[i]); dev.off()
}

The key is realizing that plot.lm (which is the method used by plot applied to a glm object, because glm is a subclass of lm and doesn't have its own particular plot method) displays diagnostic plots based on the which argument, and the default value of which is the same as wplot above. So basically: figure out how to create the individual sub-plots.