Hi!
In R 3.0.0 from current SVN, ?summary.lm says:
> Value [...]
> df degrees of freedom, a 3-vector (p, n-p, p*), the last
> being the number of non-aliased coefficients.
?summary.glm says:
> df a 3-vector of the rank of the model and the number of residual
> degrees of freedom, plus number of non-aliased coefficients.
It seems to me that the description is reversed: p is the number of
non-aliased coefficients, and p* the total number of coefficients. I do
not have reference books off-hand to check how it is intended to work,
but see this example:
ctl <- c(4.17,5.58,5.18,6.11,4.50,4.61,5.17,4.53,5.33,5.14)
trt <- c(4.81,4.17,4.41,3.59,5.87,3.83,6.03,4.89,4.32,4.69)
group <- gl(2,10,20, labels=c("Ctl","Trt"))
weight <- c(ctl, trt)
lm.D9 <- lm(weight ~ group + I(group != "Ctl"))
lm.D9
Call:
lm(formula = weight ~ group + I(group != "Ctl"))
Coefficients:
(Intercept) groupTrt I(group != "Ctl")TRUE
5.032 -0.371 NA
summary(lm.D9)$df
[1] 2 18 3
sum(!summary(lm.D9)$aliased)
[1] 2
The same is true with glm().
Also, ?summary.lm seems to miss a mention that is present
in ?summary.glm:
> Aliased coefficients are omitted in the returned object but (as from
> R 1.8.0) restored by the print method.
This is apparently true of summary.lm too:
summary(lm.D9)
Call:
lm(formula = weight ~ group + I(group != "Ctl"))
Residuals:
Min 1Q Median 3Q Max
-1.0710 -0.4938 0.0685 0.2462 1.3690
Coefficients: (1 not defined because of singularities)
Estimate Std. Error t value Pr(>|t|)
(Intercept) 5.0320 0.2202 22.850 9.55e-15 ***
groupTrt -0.3710 0.3114 -1.191 0.249
I(group != "Ctl")TRUE NA NA NA NA
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 0.6964 on 18 degrees of freedom
Multiple R-squared: 0.07308, Adjusted R-squared: 0.02158
F-statistic: 1.419 on 1 and 18 DF, p-value: 0.249
summary(lm.D9)$coefficients
Estimate Std. Error t value Pr(>|t|)
(Intercept) 5.032 0.2202177 22.85012 9.547128e-15
groupTrt -0.371 0.3114349 -1.19126 2.490232e-01
Attached is a patch that applies these changes, if I'm not mistaken (and
my English can be improved...).
Regards
Index: src/library/stats/man/summary.glm.Rd
===================================================================
--- src/library/stats/man/summary.glm.Rd (révision 62006)
+++ src/library/stats/man/summary.glm.Rd (copie de travail)
@@ -89,7 +89,8 @@
\item{dispersion}{either the supplied argument or the inferred/estimated
dispersion if the latter is \code{NULL}.}
\item{df}{a 3-vector of the rank of the model and the number of
- residual degrees of freedom, plus number of non-aliased coefficients.}
+ residual degrees of freedom, plus number of coefficients (including
+ aliased ones).}
\item{cov.unscaled}{the unscaled (\code{dispersion = 1}) estimated covariance
matrix of the estimated coefficients.}
\item{cov.scaled}{ditto, scaled by \code{dispersion}.}
Index: src/library/stats/man/summary.lm.Rd
===================================================================
--- src/library/stats/man/summary.lm.Rd (révision 62006)
+++ src/library/stats/man/summary.lm.Rd (copie de travail)
@@ -37,6 +37,9 @@
coefficients, standard errors, etc. and additionally gives
\sQuote{significance stars} if \code{signif.stars} is \code{TRUE}.
+ Aliased coefficients are omitted in the returned object but restored
+ by the \code{print} method.
+
Correlations are printed to two decimal places (or symbolically): to
see the actual correlations print \code{summary(object)$correlation}
directly.
@@ -58,8 +61,9 @@
error
\deqn{\hat\sigma^2 = \frac{1}{n-p}\sum_i{w_i R_i^2},}{\sigma^2 = 1/(n-p) Sum(w[i] R[i]^2),}
where \eqn{R_i}{R[i]} is the \eqn{i}-th residual, \code{residuals[i]}.}
- \item{df}{degrees of freedom, a 3-vector \eqn{(p, n-p, p*)}, the last
- being the number of non-aliased coefficients.}
+ \item{df}{degrees of freedom, a 3-vector \eqn{(p, n-p, p*)}, the first
+ being the number of non-aliased coefficients, the last being the total
+ number of coefficients.}
\item{fstatistic}{(for models including non-intercept terms)
a 3-vector with the value of the F-statistic with
its numerator and denominator degrees of freedom.}
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
