The following is a list of known errors in the book. -------------------- 3 The example 5 lines from the bottom should be: plot(skips ~ Opening + Mask) -------------------- 281 line 4. The call to step.gam should be step.gam(object, scope, scale, direction, trace = T, keep, steps, ...) -------------------- 292 pch=183 not a general plot character, only for postscript() -------------------- 301 1st line. need gam.wlist<-"lo" above the call to gam to get this printout -------------------- 441 1st line. Object `Untreated' never used. Omit this line (or repeat the line on page 430 defining `Treated'). -------------------- ****************************************************************** The following are found in the first printing only. Current copies (2nd printing on) should be ok. Some ``errors'' are actually fixes to the software since the book was printed. 25 7th line. There should be a period after "evaluated". -------------------- 67 The example says car.summary, meant cu.summary > car.test.frame <- cu.summary[ok,] -------------------- 72 The plot in Figure 3.3 has different order than the call, which should be: > pairs(~Mileage + Disp. + Weight) -------------------- 87 Should not mention an object .Search.list (defunct) but talk only about the search list as an internal construction in the evaluator. -------------------- 92-93 The example using bs() reverts to talking about poly() at the bottom of page 92. -------------------- 130 Need to make the dimensions of coef's right in the example at the bottom of the page. Following from Ron Thisted is ok until S has a more elegant way of making matrix out of a row: bi <- t(coef(lmf) - t(coef(lmi))) # in the main text and bi <- t(coef(fit) - t(coef(lmi))) # in the function dfbetas. -------------------- 160 summary(praov) now prints estimated variances as the 4th column of the table, not an extra line at the bottom. -------------------- 180 The reported max. abs. corr. should be .184 (error in alias.lm()) -------------------- 183 The dimnames(gunproj1)[[2]] reflect a broken naming algorithm for nested terms at the time of the book printing. The correct result is as follows: > dimnames(gunproj1)[[2]] [1] "(Intercept)" "Method" [3] "Physique.L" "Physique.Q" [5] "PhysiqueSTeam1" "PhysiqueATeam1" [7] "PhysiqueHTeam1" "PhysiqueSTeam2" [9] "PhysiqueATeam2" "PhysiqueHTeam2" [11] "Residuals" -------------------- 233 6 lines from the end of Cook.terms. XWX does not exist, should be I; also an unecessary ) in same line -------------------- 256 need plot(auto.add,residuals=T) to produce figure 7.2 -------------------- 257 pick=="ATT" in figure 7.4 -------------------- 270 names(kyph.gam1) looks slightly different now -------------------- 273 top of page delete ",degree = 1" -------------------- 289 7 lines from the bottom "fit" should read "fitted" -------------------- 291 missing quotes around "response" 3 lines from bottom of page -------------------- 301 Example using the "air" data.frame calls data named "rad" and "temp". These should be "radiation" and "temperature" respectively. -------------------- 353 names `ew.jittered', `ns.jittered' lose the `ed' in later use. -------------------- 470 In the middle of the first paragraph of page 470 are two repeated words: arguments override rguments override any corresponding arguments -------------------- 479 `match()' should be added to table A.2 `compname()', log10(), and "**"() should be removed. -------------------- 516 control.gam() should be gam.control() -------------------- 519 control.gam() should be glm.control() -------------------- 537 "See nls.control ..." but don't call it (a la the ms documentation) -------------------- 552 plot.tree(x, not plot(x, -------------------- 575 n<50 (at bottom of page)