[R] partial residuals & the output of residuals.lm(..., type="partial")

[Daniel McGlinn]

Dear list,

I would like to know how the function residuals.lm calculates the 
partial residuals from an lm object with more than one predictor 
variable.  In other words what is residuals.lm(...,type="partial") doing 
behind the scenes?  According to the help file for residuals.lm 
(?residuals.lm), "The partial residuals are a matrix with each column 
formed by omitting a term from the model".  Unfortunately, I cannot seem 
to recreate the results of the function "residuals.lm" by simply 
dropping a variable from a model and then calculating the raw residuals 
of the updated model.  Can anyone see what I am overlooking?  It may be 
helpful to others if I mention that the usage of 
residuals.lm(...,type='partial') by the function termplot is what 
motivated me to look at this function more closely.  Below is a simple 
example to illustrate my question:

set.seed(12)

x1 <- runif(100)
x2 <- runif(100)

y <- .13+.25*x1+.70*x2+runif(100)

mod <- lm(y~x1+x2)

##let's only consider the partial residuals when x2 is dropped from the 
model
plot(residuals(mod,type="partial")[,2],residuals(update(mod,.~.-x2),type='response'))
abline(0,1) ##1:1 line
##why do the points not all fall on the 1:1 line?

Thanks,
Dan

-- 
Daniel J. McGlinn
Department of Botany, Oklahoma State University
http://ecology.okstate.edu/Libra/