[R] Question on Plotting

MessageI believe there are several packages that can do this. I have found
the errorbars function in Frank Harrell's Hmisc
package (http://hesweb1.med.virginia.edu/biostat/s/Hmisc.html) to be quite
good.

There is an example in ?predict.lm (which you should read), but it
does not include the scatterplot. So if you want the data points
themselves included in your graph, you can use something like the
following:

# Generate x and y
x <- rnorm(50)
y <- x + rnorm(50)

# Create model
mod <- lm(y ~ x)

# Generate new dataframe containing prediction data
# of 50 points over the range of x values
# This does two things:
# 1. It limits the prediction based values of x to
# the range of known data
# 2. It keeps the values of x in numeric sequence for
# matlines()
# Be sure that the new data frame colnames are the same
# as your model terms
new <- data.frame(x = seq(min(x), max(x),
by = (max(x) - min(x)) / 49))

# Now create matrix of fitted values and upper and lower limits
# for prediction bands and confidence bands
pred.lim <- predict(mod, newdata = new, interval = "prediction")
conf.lim <- predict(mod, newdata = new, interval = "confidence")

# Generate plot, setting y axis to include range of
# prediction values
plot(x, y, xlab = "x", ylab = "y", col = "black", bg = "red",
pch = 21, las = 1, cex = 0.9, ylim = range(y, pred.lim, na.rm =
TRUE))

# Use matlines() to draw fitted line and pred/conf bands
matlines(new$x, pred.lim, lty = c(1, 4, 4),
lwd = c(2, 1, 1), col = c("black", "red", "red"))

matlines(new$x, conf.lim, lty = c(1, 3, 3),
lwd = c(2, 1, 1), col = c("black", "darkgreen", "darkgreen"))



Hope that helps,

Marc Schwartz