panel.box: Panel functions for predicted values and SE box

In gmonette/spida2: Collection of tools developed for the Summer Programme in Data Analysis 2000-2012

Panel functions for predicted values and SE box

Description

With 'layer' and 'glayer' in 'latticeExtra', these functions can be used to easily generate fitted values and error boxes that have a reasonable appearance whether a plot uses 'groups' or not.

Usage

panel.box(
  x,
  y,
  se,
  lower = y - se,
  upper = y + se,
  width = if (is.numeric(x)) max(diff(sort(x)))/2 else 1/2,
  subscripts,
  ...,
  type,
  group.number,
  alpha = 0.9,
  alpha.fit = alpha,
  alpha.box = 0.3,
  col,
  col.line,
  col.symbol,
  border = F,
  font,
  fontface
)

Arguments

x, y	arguments passed by 'layer'
lower, upper	FIXME
subscripts	FIXME
group.number	FIXME
alpha	FIXME
col	FIXME
col.symbol	is used to control color when using 'groups'
border	default = FALSE for panel.band.
font	FIXME
fontface	FIXME
data	data frame to be used to add additional values of numeric variable
form	formula evaluated in data. The first term defines the variable with values to be filled in and the remaining terms define the variables to be used for grouping determining the minima and maxima within which values are added.
xpd	expansion factor to add points beyond minima and maxima. Default 1.0.
fit	fitted values of a model, generally passed through 'layer' from a call to 'xyplot': e.g. xyplot( y ~ x, data, groups = g, fit = data$yhat, lower = with(data, yhat - 2*se), upper = with(data, yhat + 2*se), subscripts = T)
dots	FIXME

Value

The 'panel.bands', 'panel.fit', and 'panel.labels' functions are invoked for their graphical effect.

Author(s)

Georges Monette [e-mail]

Examples

## Not run: 
  library(spida2)
  library(latticeExtra)
  library(car)
  Prestige$Education <- cut(Prestige$education,
         breaks = c(-Inf,8,12,+Inf), labels = c('< HS','some HS','some PS'))
  Prestige$Gender <- cut(Prestige$women, breaks = c(-Inf, 10, 50, +Inf),
         labels = c('Male','Mixed','Female'))

  fit <- lm(income ~ Gender * Education, Prestige,
      na.action = na.exclude)
  pred <- with(Prestige, expand.grid(Education = levels(Education),
           Gender = levels(Gender)))
  pred <- cbind(pred, predict(fit, newdata = pred, se = TRUE))
  pred
  (p <- xyplot( fit ~ Education | Gender , pred,  
                subscripts = T, width = .5,
                fit = pred$fit,
                labels = pred$Education,
                lower = with(pred, fit - 2*se.fit),
                upper = with(pred, fit + 2*se.fit)))
  gd(3,pch = 16)
  p + layer(panel.fit(...))
  (p <- update(p + layer(panel.box(...)), ylim = c(0, 15000), 
         auto.key = list(columns= 3)))
  ###### Need to fix handling of width with factor predictor   
  p + layer(panel.labels(...))
  
# NOTE: with groups use 'glayer' instead of 'layer' 

## End(Not run)

gmonette/spida2 documentation built on Aug. 20, 2023, 7:21 p.m.