visreg
An R package for the visualization of regression models

visreg tries to set up pleasant-looking default options, but everything can be tailored to user specifications. For the plots below, we work from this general model:

airquality$Heat <- cut(airquality$Temp, 3, labels=c("Cool", "Mild", "Hot"))
fit <- lm(Ozone ~ Solar.R + Wind + Heat, data=airquality)

Turning on/off plot components

By default, visreg includes the fitted line, confidence bands, and partial residuals, but the residuals and the bands can be turned off:

visreg(fit, "Wind", band=FALSE)

plot of chunk noband

visreg(fit, "Wind", partial=FALSE)

plot of chunk nopartial

Note that by default, when you turn off partial residuals, visreg tries to display a rug so you can at least see where the observations are. You can turn this off too:

visreg(fit, "Wind", partial=FALSE, rug=FALSE)

plot of chunk nopartial_norug

Finally, there is an option for displaying separate rugs for positive and negative residuals on the top and bottom axes, respectively, with rug=2 (this is particularly useful for logistic regression):

visreg(fit, "Wind", rug=2, partial=FALSE)

plot of chunk rug2

Jittering

If there are many ties in a numeric variable x, jittering can be helpful way to avoid overplotting:

fit <- lm(Ozone ~ Solar.R + Wind + poly(Month, 2), data=airquality)
visreg(fit, "Month")

plot of chunk nojit

visreg(fit, "Month", jitter=TRUE)

plot of chunk jit

Appearance of points, lines, and bands

Specifying col='red' won’t work, because visreg can’t know whether you’re trying to change the color of the line, the band, or the points. These options must be specified through separate parameters lists:

Each of these can be abbreviated, as in the example below:

visreg(fit, "Wind", line=list(col="red"),
                    fill=list(col="green"),
                    points=list(cex=1.5, pch=1))

plot of chunk aes

Generic plot options

Other options get passed along to plot, so any option that you could normally pass to plot, like main, will work fine. Here’s an example that includes a bunch of options like this:

fit <- lm(log(Ozone) ~ Solar.R + Wind + Temp, data=airquality)
visreg(fit, "Wind", yaxt="n", main="Ozone is bad for you", bty="n", ylab="Ozone")
at <- seq(1.5, 5, 0.5)
lab <- round(exp(at), 1)
axis(2, at=at, lab=lab, las=1)

plot of chunk generic

Whitespace for factors

When x is a factor, the whitespace option controls the amount of whitespace in between the categories:

fit <- lm(Ozone ~ Solar.R + Wind + Heat, data=airquality)
par(mfrow=c(1,2))
visreg(fit, "Heat", whitespace=.1)
visreg(fit, "Heat", whitespace=.5)

plot of chunk whitespace

Subsetting the plot

Occasionally, you might want to plot only a subset of the levels or observations; you can use subset to accomplish this:

v <- visreg(fit, "Wind", by="Heat", plot=FALSE)
v1 <- subset(v, Heat %in% c("Cool", "Hot"))
plot(v1)

plot of chunk subset_1

v2 <- subset(v, Wind < 15)
plot(v2, layout=c(3,1))

plot of chunk subset_2