The R Book

Serial correlation in the residuals

The Durbin-Watson function is used for testing whether there is autocorrelation in the residuals from a linear model or a GLM, and is implemented as part of the car package (see Fox, 2002):

install.packages("car")
model<-lm(response~explanatory)
durbin.watson(model)

lag  Autocorrelation  D-W Statistic p-value
  1      -0.07946739       2.049899   0.874
Alternative hypothesis: rho != 0

There is no evidence of serial correlation in these residuals (p = 0.874).

The car package also contains functions for drawing ellipses, including data ellipses, and confidence ellipses for linear and generalized linear models. Here is the data.ellipse function for the present example: by default, the ellipses are drawn at 50 and 90%:

data.ellipse(explanatory,response)