The R Project SVN R

Rev

Rev 77208 | Rev 84026 | Go to most recent revision | Show entire file | Ignore whitespace | Details | Blame | Last modification | View Log | RSS feed

Rev 77208 Rev 77213
Line 29... Line 29...
29 
}
 29 
}
30 
\details{
 30 
\details{
31 
 The \pkg{stats} package provides the S3 generic and a default method.
 31 
 The \pkg{stats} package provides the S3 generic and a default method.
32 
 The latter is correct typically for (asymptotically / approximately)
 32 
 The latter is correct typically for (asymptotically / approximately)
33 
 generalized gaussian (\dQuote{least squares}) problems, since it is
 33 
 generalized gaussian (\dQuote{least squares}) problems, since it is
- 
 
 34 
 defined as
34 
 defined as \preformatted{%
 35 
 \preformatted{
35 
   sigma.default <- function (object, use.fallback = TRUE, ...)
 36 
   sigma.default <- function (object, use.fallback = TRUE, ...)
36 
                      sqrt( deviance(object, ...) / (NN - PP) )
 37 
       sqrt( deviance(object, ...) / (NN - PP) )
- 
 
 38 
 }
37 
 } where \code{NN <- \link{nobs}(object, use.fallback = use.fallback)}
 39 
 where \code{NN <- \link{nobs}(object, use.fallback = use.fallback)}
38 
 and \code{PP <- sum(!is.na(\link{coef}(object)))} -- where in older \R
 40 
 and \code{PP <- sum(!is.na(\link{coef}(object)))} -- where in older \R
39 
 versions this was \code{length(coef(object))} which is too large in
 41 
 versions this was \code{length(coef(object))} which is too large in
40 
 case of undetermined coefficients, e.g., for rank deficient model fits.
 42 
 case of undetermined coefficients, e.g., for rank deficient model fits.
41 
}
 43 
}
42 
\value{
 44 
\value{