Pure Error analysis of variance

Description

For a linear model object of type lm, finds the sum of squares for lack of fit and the sum of squares for pure error. These are added to the standard anova table to give a test for lack of fit. If there is no pure error, then the regular anova table is returned.

Usage

pure.error.anova(mod)

Arguments

Details

For regression models with one predictor, say y ~ x, this method fits y ~ x + factor(x) and prints the anova table. With more than one predictor, it computes a random linear combination L of the terms in the mean function and then gives the anova table for update(mod, ~.+factor(L)).

Value

Returns an analsis of variance table.

Author(s)

Sanford Weisberg, sandy@stat.umn.edu

References

Weisberg, S. (2005). Applied Linear Regression, third edition, New York: Wiley, Chapter 5.

See Also

lm

Examples

x <- c(1,1,1,2,3,3,4,4,4,4)
y <- c(2.55,2.75,2.57,2.40,4.19,4.70,3.81,4.87,2.93,4.52)
m1 <- lm(y~x)
anova(m1)  # ignore pure error
pure.error.anova(m1)  # include pure error

data(forbes)
m2 <- lm(Lpres~Temp, data=forbes)
pure.error.anova(m2)  # function does nothing because there is no pure error