[R] lm(y ~ x1) vs. lm(y ~ x0 + x1 - 1) with x0 <- rep(1, length(y))

[jochen laubrock]

Dear list,

the following came up in an introductory class. Please help me understand the -1 (or 0+) syntax in formulae: Why do the enumerator dfs, F-statisics etc. differ between the models lm(y ~ x1) and lm(y ~ x0 + x1 - 1), if x0 is a vector containing simply ones?

Example:

N  <- 40
x0 <- rep(1,N)
x1 <- 1:N
vare <- N/8
set.seed(4)
e <- rnorm(N, 0, vare^2)

X <- cbind(x0, x1)
beta <- c(.4, 1)
y <- X %*% beta + e

summary(lm(y ~ x1))
# [...]
# Residual standard error: 20.92 on 38 degrees of freedom
# Multiple R-squared: 0.1151,	Adjusted R-squared: 0.09182 
# F-statistic: 4.943 on 1 and 38 DF,  p-value: 0.03222 

summary(lm(y ~ x0 + x1 - 1))        # or summary(lm(y ~ 0 + x0 + x1))
# [...]
# Residual standard error: 20.92 on 38 degrees of freedom
# Multiple R-squared: 0.6888,	Adjusted R-squared: 0.6724
# F-statistic: 42.05 on 2 and 38 DF,  p-value: 2.338e-10 


Thanks in advance,
Jochen


----
Jochen Laubrock, Dept. of Psychology, University of Potsdam,
Karl-Liebknecht-Strasse 24-25, 14476 Potsdam, Germany
phone: +49-331-977-2346, fax: +49-331-977-2793