Skip to content
Prev 276983 / 398506 Next

ggplot2 - regression statistics how to display on plot

Hi:

Here's an example of how one might do this in a specific example using
geom_text().

# Some fake data:
df <- data.frame(x = 1:10, y = 0.5 + (1:10) + rnorm(10))

# Fit a linear model to the data and save the model object:
mod <- lm(y ~ x, data = df)

# Create a list of character strings - the first component
# produces the fitted model, the second produces a
# string to compute R^2, but in plotmath syntax.

rout <- list(paste('Fitted model: ', round(coef(mod)[1], 3), ' + ',
                               round(coef(mod)[2], 3), ' x', sep = ''),
              paste('R^2 == ', round(summary(mod)[['r.squared']], 3),
sep = '')  )

# This looks ugly, but I'm using the round() function to make the
# equations look more sane. coef(mod) extracts the model
# coefficients (intercept, then slope), and
# summary(mod)[['r.squared']] extracts R^2.

# See what they look like:
rout

# Notice that the first component of rout is simply a text string
# that can be passed as is, but the second string needs to be
# wrapped inside an expression, which is what parse = TRUE
# in geom_text() does.

# Now construct the plot; given the (x, y) extent of the data, the
# coordinates make sense, but you have to adapt them to
# your data.

ggplot(df, aes(x, y)) +
   geom_smooth(method = 'lm') +
   geom_text(aes(x = 2, y = 10, label = rout[[1]]), hjust = 0) +
   geom_text(aes(x = 2, y = 9.5, label = rout[[2]]), hjust = 0, parse = TRUE )

hjust = 0 makes the text strings flush left, parse = TRUE in the
second call converts the second string in rout into an expression.

This is fine if your plot is a one-off or two-off deal; the code above
gives you a sense of the programming required. OTOH, if you need to do
this a lot, then you're better off with a function that can automate
the process, which Bryan was kind enough to provide.

I might add that there is a dedicated ggplot2 listserv on google:
ggplot2 at googlegroups.com, for which questions like these are well
suited.

HTH,
Dennis

On Thu, Nov 10, 2011 at 5:45 AM, Durant, James T. (ATSDR/DTEM/PRMSB)
<hzd3 at cdc.gov> wrote: