plotEst: Plot estimates with confidence limits (forest plot)

Description

Plots parameter estimates with confidence intervals, annotated with parameter names. A dot is plotted at the estimate and a horizontal line extending from the lower to the upper limit is superimposed.

Usage

plotEst( ests, y = dim(ests)[1]:1, txt = rownames(ests), txtpos = y, ylim = range(y)-c(0.5,0), xlab = "", xtic = nice(ests[!is.na(ests)], log = xlog), xlim = range( xtic ), xlog = FALSE, pch = 16, cex = 1, lwd = 2, col = "black", col.txt = "black", font.txt = 1, col.lines = col, col.points = col, vref = NULL, grid = FALSE, col.grid = gray(0.9), restore.par = TRUE, ... )
linesEst( ests, y = dim(ests)[1]:1, pch = 16, cex = 1, lwd = 2, col="black", col.lines=col, col.points=col, ... )
pointsEst( ests, y = dim(ests)[1]:1, pch = 16, cex = 1, lwd = 2, col="black", col.lines=col, col.points=col, ... )

Arguments

ests

Matrix with three columns: Estimate, lower limit, upper limit. If a model object is supplied, ci.lin is invoked for this object first.

Vertical position of the lines.

txt

Annotation of the estimates. Either a character vector or an expression vector.

txtpos

Vertical position of the text. Defaults to y.

ylim

Extent of the vertical axis.

xlab

Annotation of the horizontal axis.

xtic

Location of tickmarks on the x-axis.

xlim

Extent of the x-axis.

xlog

Should the x-axis be logarithmic?

pch

What symbol should be used?

cex

Expansion of the symbol.

col

Colour of the points and lines.

col.txt

Colour of the text annotating the estimates.

font.txt

Font for the text annotating the estimates.

col.lines

Colour of the lines.

col.points

Colour of the symbol.

lwd

Thickness of the lines.

vref

Where should vertical reference line(s) be drawn?

grid

If TRUE, vertical gridlines are drawn at the tickmarks. If a numerical vector is given vertical lines are drawn at grid.

col.grid

Colour of the vertical gridlines

restore.par

Should the graphics parameters be restored? If set to FALSE the coordinate system will still be available for additional plotting, and par("mai") will still have the very large value set in order to make room for the labelling of the estimates.

...

Arguments passed on to ci.lin when a model object is supplied as ests.

Value

NULL

Details

plotEst makes a news plot, whereas linesEst and pointsEst (identical functions) adds to an existing plot.

If a model object of class "glm", "coxph", "clogistic" or "gnlm" is supplied the argument xlog defaults to TRUE, and exponentiated estimates are extracted by default.

Examples

Run this code

# Bogus data and a linear model
f <- factor( sample( letters[1:5], 100, replace=TRUE ) )
x <- rnorm( 100 )
y <- 5 + 2 * as.integer( f ) + 0.8 * x + rnorm(100) * 2
m1 <- lm( y ~ f )

# Produce some confidence intervals for contrast to first level
( cf <- ci.lin( m1, subset=-1 )[,-(2:4)] )

# Plots with increasing amounts of bells and whistles
par( mfcol=c(3,2), mar=c(3,3,2,1) )
plotEst( cf )
plotEst( cf, grid=TRUE, cex=2, lwd=3 )
plotEst( cf, grid=TRUE, cex=2, col.points="red", col.lines="green" )
plotEst( cf, grid=TRUE, cex=2, col.points="red", col.lines="green",
             xlog=TRUE, xtic=c(1:8), xlim=c(0.8,6) )
rownames( cf )[1] <- "Contrast to fa:\n fb"
plotEst( cf, grid=TRUE, cex=2, col.points=rainbow(4),
                                col.lines=rainbow(4), vref=1 )
#
etxt <- expression("Plain text, qouted",
                   "combined with maths:"*sqrt(a)*phi[c],
                   f^d*"  Hb"*A[1][c],
                   eff^e*"  kg/"*m^2)
plotEst( cf, txt=etxt, grid=TRUE, cex=2, col.points=rainbow(4),
                                         col.lines =rainbow(4), vref=1 )