`Modern Applied Statistics with S'
==================================
by W.N. Venables & B.D. Ripley
==============================

Errata and comments for the first printing of the fourth edition.

p. 15 In R there is no class "named" and so the named vector still has
      class "numeric".

p. 19 Third and fourth lines of text, omit >...< in:

    Matrices and arrays can also have names for the dimensions, known as
    dimnames.  The simple way to add them is >to< just to assign them,
    using NULL where we do not want >a< to specify a set of names.


p. 41 The use of _ for assignment is deprecated in S-PLUS 6.1 and in R
      (from 1.6.0): it will be removed in R 1.8.0.

p. 78 l-2 (xl, xu, yl, yu), not (xl, xu, yl, yl)
       R does not have prompt.screen.

p. 83 In R, you need par(new=TRUE) before the second par(fig=) call.

p.100 The background to this figure (Figure 3 of the reference) can be
      found in 

    Bradley, K.M., Bydder, G.M., Budge, M.M., Hajnal, J.V., White, S.J.,
    Ripley, B.D. and Smith, A.D. (2002) Serial brain MRI at 3-6 month
    intervals as a surrogate marker for Alzheimer's disease.
    British Journal of Radiology 75, 506-513.

p.162 The version of rlm prior to May 2004 used case weights, not
    inverse variance weights.  The current version gives the more
    correct results of

Coefficients:
              Value Std. Error t value
(Intercept)  -4.062   1.651     -2.461
       dist   5.822   0.452     12.889
      climb   0.008   0.001      5.937

Residual standard error: 0.819 on 32 degrees of freedom

p.190 The gaussian family in R allows links "log" and "inverse" and
      has for a long time despite the help page prior to R 1.7.0.

pp.198, 478  Santner not Santer.

p.199-200, 468  Cox and Snell (1981), not 1984.

p.203 l.-2 replace `:;' by `: '

p.210 l.-3 the>s<e are

p.222 l.-2 replace V by v for consistency with the following formulae.

p.231 (8.7) y_i should be y_j.

p.232 l.6 Rams>a<y

p.265 l.1 of text: replace `devices' by `device'.

p.272 l,12 but >the< clearly: omit >the<.

p.297 Table 10.4.  In the first column, the se for V$ should be 0.05,
	not 0.15 -- see page 294 for the actual output.

p.308 l.5 The reference to figure 11.4 should be to the top right
	panel of figure 11.3.

P.311 l.7 has a spurious `batchSOM' at the beginning of the line.

p.320-1 The authors of mclust have chosen to re-use the name for a
      completely incompatible package.  We can no longer recommend its
      use, and the code given here does not work in R's mclust-2.x.
      (The older S-PLUS version is still available at the web site
      given.)  The R scripts now give code that will work under
      mclust-2.x, but it does not give the same answers.

pp.381-2  S-PLUS 6.1 gives different numbers for `n' in the survfit
	printouts, the number of records not the number of
	individuals.  (This looks like a bug.)

p.426 The third para of the display should have trsurf replaced by
	prsurf, as in 

    prsurf <- prmat(topo.gls, 0, 6.5, 0, 6.5, 50)
    eqscplot(prsurf, , xlab = "", ylab = "", type = "n")
    contour(prsurf, levels = seq(600, 1000, 25), add = T)

The figure was computed by the correct code.

p.454 l.5 replace `put' by `puts'.

p.463 l-7.  `convenient' not `convention'.

p.473 Hyndman & Fan clearly do not understand the point of URLs and
      have made a gratuitious change to
http://www-personal.buseco.monash.edu.au/~hyndman/Rlibrary/misc/quantile.htm

p.480 O'Connell not O''Connell.



S-PLUS Changes
==============

p.416  The GARCH facilities are now available in module S+FinMetrics,
	so replace module(garch) by module(finmetrics) if you have the
	latter (from S-PLUS 6.2).


R Changes
=========

p.12  As from R 1.7.0 data() is not needed for our datasets, but it
      is needed for R's own datasets  ability.cov, iris3  and swiss
      prior to R 2.0.0.

p.13  footnote 1.  As from R 1.7.0 class() will always report a
      non-null class.

p.69  There are now two R packages called `rgl', although efforts are
      under way to merge them.  That described in the footnote is now 
      called `djmrgl'.

p.76  R's lattice package has since added a contourplot() function.

p.98  R's lattice package version >=0.6-5 requires adding

	y <- as.numeric(y)

      as the first line of this panel function.

pp.128,132,134  The new (Dec 2002) R package polspline should be used in
      place of logspline, by replacing dlogspline by doldlogspline,
      and logspline.fit by oldlogspline.  Feb 2004: the old logspline
      package has been withdrawn.

p.299 yags for R is now yags3, and the syntax has been changed at version
      3.21.  See the scripts in the R MASS package for the changes needed.

pp.307, 308 (twice) replace codes() by unclass(), as codes() has been
     withdrawn in R.  See the scripts in the R MASS package.