Wentzel-Larsen, Tore wrote: > Thanks, > I use Design version 2.1-1 (and as provided, R 2.5.1 on Windows XP).
Sorry I missed the latter. > The redundancies in object names were for my own convience, as part of > a larger command file, and the validation of this univariate model was > only included to ease comparison with the main multivariate model. I have A minor point: adding more variables to the right hand side makes the model a multivariable model. It is still univariate as it has only one dependent variable. > tried to access Design version 2.0_12, but only managed to access version > 2.1_1 of Design in my Windows implementation of R2.5.1 (the choice of > operative system is made by my institution and I am only entitled to use > Windows). My mistake again. I forgot that the Debian repositories for CRAN are behind. I updated to latest Design in CRAN and found the bug. I will send a separate note with a new version of the function in question for you to source( ) to override the current function. Frank > Best, Tore > > -----Opprinnelig melding----- > Fra: Frank E Harrell Jr [mailto:[EMAIL PROTECTED] > Sendt: 28. august 2007 03:17 > Til: Wentzel-Larsen, Tore > Kopi: r-help@stat.math.ethz.ch > Emne: Re: [R] validate (package Design): error message "subscript out of > bounds" > > Wentzel-Larsen, Tore wrote: >> Dear R users >> >> I use Windows XP, R2.5.1 (I have read the posting guide, I have >> contacted the package maintainer first, it is not homework). >> >> In a research project on renal cell carcinoma we want to compute >> Harrell's c index, with optimism correction, for a multivariate >> Cox regression and also for some univariate Cox models. >> For some of these univariate models I have encountered an error >> message (and no result produced) from the function validate i >> Frank Harrell's Design package: >> >> Error in Xb(x[, xcol, drop = FALSE], coef, non.slopes, non.slopes.in.x, : >> subscript out of bounds >> >> The following is an artificial example wherein I have been able to >> reproduce this error message (actual data has been changed to preserve >> confidentiality): > > I could not reproduce the error on R 2.5.1 on linux using version 2.0-12 > of Design (you did not provide this information). > > Your code involved a good deal of extra typing. Here is a streamlined > version: > > bc <- data.frame(time1 = c(9,24,28,43,58,62,66,107,116,118,123, > 127,129,131,137,138,139,140,148,169,176,179,188,196,210,218, > .... > bc > > library(Design) > > dd <- with(bc, datadist(bc1, age, adjto.cat='first')) > options(datadist = 'dd') > > f <- cph(Surv(time1,status1) ~ bc1, > data = bc, x=TRUE, y=TRUE, surv=TRUE) > anova(f) > f > summary(f) > > val <- validate(f, B=200, dxy=TRUE) > > I don't get much value of putting the type of an object as part of the > object's name, as information within objects defines the object type/class. > > There is little reason to validate a one degree of freedom model. > > Frank > >> library(Design) >> >> # an example data frame: >> frame.bc <- data.frame(time1 = c(9,24,28,43,58,62,66,107,116,118,123, >> 127,129,131,137,138,139,140,148,169,176,179,188,196,210,218, >> 1,1,1,2,2,3,4,8,23,32,33,34,43,44,48,51,52,54,59,59,60,60,62, >> 65,65,68,70,72,73,74,81,84,88,98,99,106,107,115,115,117,119, >> 120,122,122,122,122,126,128,130,135,136,136,138,149,151,154, >> 157,159,161,164,164,164,166,172,172,176,179,180,183,183,184, >> 187,190,197,201,201,203,203,203,209,210,214,219,227,233,4,18, >> 49,113,147,1,1,2,2,2,2,2,3,4,6,6,6,6,6,6,6,6,9,9,9,9,9,10,10, >> 10,11,12,12,12,13,14,14,17,18,18,19,19,20,20,21,21,21,21,22,23, >> 23,24,28,28,29,29,32,34,35,38,38,48,48,52,52,54,54,56,64,67,67, >> 69,70,70,72,84,88,90,114,115,140,142,154,171,195), >> status1 = c(0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, >> 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, >> 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, >> 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1,1,1,1,1,1, >> 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, >> 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, >> 1,1,1,1,1), >> bc1 = factor(c(1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, >> 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, >> 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, >> 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,1,1,1,1,1,2,2,2,2,2,2,2,2,2,2,2,2, >> 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, >> 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2), >> labels=c('bc.1','bc.2')), >> age = c(58,68,23,20,50,43,41,69,20,48,19,27,39,20,65,49,70,59,31,43,25, >> 61,60,45,34,59,32,58,30,62,26,44,52,29,40,57,33,18,50,50,55,51,38,34, >> 69,56,67,38,66,21,48,39,62,62,29,68,66,19,60,39,55,42,24,29,56,61,40, >> 52,19,40,33,67,66,51,48,63,60,58,68,60,53,20,45,62,37,38,61,63,43,67, >> 49,39,43,67,49,69,32,37,32,63,33,47,66,39,23,57,26,61,20,49,69,30,40, >> 29,38,66,60,69,69,44,65,25,41,53,18,55,45,59,49,27,51,29,67,26,24,26, >> 47,23,50,27,35,45,32,26,45,45,63,39,39,22,38,27,31,27,49,65,66,49,39, >> 21,51,49,55,63,19,26,50,21,24,34,65,33,55,33,36,53,48,25,54,58,60,34, >> 47,23,34,60,39,34,22,30,41,55,64,48,34,54)) >> frame.bc >> >> # preparing for a simple univariate Cox regression: >> dd.bc <- datadist(frame.bc[, c('bc1','age')], adjto.cat='first') >> options(datadist = 'dd.bc') >> >> # a univariate Cox regression: >> cph.bc <- cph(formula = Surv(time1,status1)~bc1, >> data = frame.bc, x=TRUE, y=TRUE, surv=TRUE) >> anova(cph.bc) >> cph.bc >> summary(cph.bc) >> >> # the validate command for the Cox model: >> val.cph.bc <- validate(cph.bc, B=200, dxy=TRUE , pr=TRUE) >> >> ---------------------- >> Output from the validate command: >> >> training test >> Dxy -0.124360 -0.1423409 >> R2 1.000000 1.0000000 >> Slope 1.000000 0.7919584 >> D 0.016791 0.0147536 >> U -0.002395 0.0006448 >> Q 0.019186 0.0141088 >> training test >> Dxy -0.191875 -0.1423409 >> R2 1.000000 1.0000000 >> Slope 1.000000 0.8936724 >> D 0.022397 0.0147536 >> U -0.002339 0.0001367 >> Q 0.024736 0.0146169 >> training test >> Dxy -0.199514 -0.1423409 >> R2 1.000000 1.0000000 >> Slope 1.000000 0.8075246 >> D 0.025717 0.0147536 >> U -0.002447 0.0005348 >> Q 0.028163 0.0142188 >> Error in Xb(x[, xcol, drop = FALSE], coef, non.slopes, non.slopes.in.x, : >> subscript out of bounds >> >> >> Any help/suggestions will be highly appreciated. >> >> >> Sincerely, >> Tore Wentzel-Larsen >> statistician >> Centre for Clinical research >> Armauer Hansen house >> Haukeland University Hospital >> N-5021 Bergen >> tlf +47 55 97 55 39 (a) >> faks +47 55 97 60 88 (a) >> email [EMAIL PROTECTED] >> >> ______________________________________________ >> R-help@stat.math.ethz.ch mailing list >> https://stat.ethz.ch/mailman/listinfo/r-help >> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html >> and provide commented, minimal, self-contained, reproducible code. >> > > -- Frank E Harrell Jr Professor and Chair School of Medicine Department of Biostatistics Vanderbilt University ______________________________________________ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
