residVSfitted

0th

Percentile

Draw plots of resid ~ y.hat and sqrt(abs(resid)) ~ y.hat

Draw plots of resid ~ y.hat and sqrt(abs(resid)) ~ y.hat. This is a pair of lattice functions that duplicate the first and third panels of stats:::plot.lm.

Keywords
hplot
Usage
residVSfitted(linearmodel, groups = (e >= 0), ...)
scaleLocation(linearmodel, groups = (e >= 0), ...)
Arguments
linearmodel
"lm" object.
groups
This is the standard groups argument for xyplot. The default value is one symbol and color for positive residuals and a different symbol and color for negative residuals.
...
Additional arguments to xyplot.
Value

  • "trellis" object.

Aliases
  • residVSfitted
  • scaleLocation
Examples
data(fat)
fat.lm <- lm(bodyfat ~ abdomin, data=fat)

A <- residVSfitted(fat.lm, pch=c(25,24),
                   fill=trellis.par.get("superpose.symbol")$col[1:2])
B <- scaleLocation(fat.lm, pch=c(25,24),
                   fill=trellis.par.get("superpose.symbol")$col[1:2])
BA <- c("Scale-Location"=B,
        "Residuals vs Fitted"=update(A, scales=list(y=list(at=-100, alternating=3))),
        layout=c(1,2))
BA

BAu <-
  update(BA,
         ylab=c(B$ylab, A$ylab),
         ylab.right=c(B$ylab.right, A$ylab.right),
         xlab.top=NULL,
         between=list(y=1),
         par.settings=list(layout.widths=list(ylab.right=6))
         )

C <- diagQQ(fat.lm)

D <- diagplot5new(fat.lm)

print(BAu, split=c(1,1,2,1), more=TRUE)

print(update(c("Normal Q-Q"=C), xlab.top=NULL, strip=TRUE),
      ## split=c(2,1,2,2),
      position=c(.5, .54, 1, 1),  ## .54 is function of device and size
      more=TRUE)

print(update(D, xlab.top=NULL,
             strip=strip.custom(factor.levels=D$xlab.top),
             par.strip.text=list(lines=1.3)),
      ## split=c(2,2,2,2),
      position=c(.5, 0, 1, .57),  ## .57 is function of device and size
      more=FALSE)
## the .54 and .57 work nicely with the default quartz window on Mac OS X.
Documentation reproduced from package HH, version 3.1-25, License: GPL (>= 2)

Community examples

Looks like there are no examples yet.