[R] Passing arguments to panels in trellis plots
Dear all, I am trying to produce survfit plots in a trellis environment and I would like the plots to be logarithmic. I am trying this: print(Ecdf(~time | size*type, groups=alg,data=B,subscripts=TRUE, panel=function(x,groups,subscripts) { t - survfit(Surv(time[subscripts],event[subscripts])~groups[subscripts],data=B) panel.xyplot(t[1]$time,1-t[1]$ssurv,type=s,lty=2) panel.xyplot(t[2]$time,1-t[2]$ssurv,type=s,lty=2) }, scale=list(log=TRUE) ) but data are transformed in logarithm before being passed to the panel and hence the output of the function survfit is not the expected one. Is there a way to plot this correctly, ie, having first the survfit computed and then the plot, like in: plot(survfit(Surv(time,event)~groups,data=B),log=true) Thanks in advance. - Marco. -- Marco Chiarandini http://www.imada.sdu.dk/~marco Department of Mathematics Email: marco AT imada.sdu.dk and Computer Science, Phone: +45 6550 4031 University of Southern DenmarkFax: +45 6593 2691 __ 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.
Re: [R] Passing arguments to panels in trellis plots
Dear Deepayan, Since you haven't bothered to follow the posting guide at all, it took me a while to figure out that you live in that alternate R universe created by Prof Harrell. Can't help you there, but in the standard universe, things seem fairly simple: I am sorry for not having been clear. I hope this time I get closer to the correct post practice. Here is what I am trying to do: D - expand.grid(rep=c(1:10),b=c(M,N),a=c(10,20),alg=c(A,B)) D1 - data.frame(D[,c(2,3,4)],time=runif(80,1,100)) D2 - data.frame(D1,event=rbinom(80,1,0.9)) library(survival) library(lattice) library(Hmisc) Ecdf(~time | a*b, groups=alg,data=D2, subscripts=TRUE, panel=function(x,groups,subscripts) { t - survfit(Surv(time[subscripts],event[subscripts])~groups[subscripts], data=D2,type=kaplan-meier, conf.type=plain,conf.int=.95, se.fit=T) panel.xyplot(t[2]$time,1-t[2]$surv,type=s,lty=2) panel.xyplot(t[1]$time,1-t[1]$surv,type=s,lty=1) } ) Although not very elegant this does the job. Nevertheless, when I try to add: scales = list(log=TRUE) in the Ecdf fucntion above I incurr in problems because the transformation of data occurs before applying the function Surv. Hence my question, is there a way to plot survfit in trellis plots with different strata (the alg factor above) in the same plot and conditional to combination of other factors (the a and b factors above)? Here is my sessionInfo(): R version 2.4.0 (2006-10-03) i686-pc-linux-gnu locale: LC_CTYPE=en_DK.UTF-8;LC_NUMERIC=C;LC_TIME=C;LC_COLLATE=C;LC_MONETARY=C;LC_MESSAGES=C;LC_PAPER=C;LC_NAME=C;LC_ADDRESS=C;LC_TELEPHONE=C;LC_MEASUREMENT=C;LC_IDENTIFICATION=C attached base packages: [1] grid splines methods stats graphics grDevices [7] utils datasets base other attached packages: Hmisc lattice survival 3.1-2 0.14-9 2.29 Regards, - Marco. -- Marco Chiarandini http://www.imada.sdu.dk/~marco Department of Mathematics Email: [EMAIL PROTECTED] and Computer Science, Phone: +45 6550 4031 University of Southern DenmarkFax: +45 6593 2691 __ 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.
[R] Survfit plots in trellis graphics
Dear all, is there a way to produce survfit plots in a trellis environment? I am trying this: print(Ecdf(~time | size*type, groups=alg,data=B,subscripts=TRUE, panel=function(x,groups,subscripts) { t - survfit(Surv(time[subscripts],event[subscripts])~groups[subscripts],data=B,conf.type=none) panel.xyplot(t[1]$time,1-t[1]$ssurv,type=s,lty=2) panel.xyplot(t[2]$time,1-t[2]$ssurv,type=s,lty=2) } ) but things get messed up if I try log transformations and to plot confidence intervals. - Marco. -- Marco Chiarandini http://www.imada.sdu.dk/~marco Department of Mathematics Email: marco AT imada.sdu.dk and Computer Science, Phone: +45 6550 4031 University of Southern DenmarkFax: +45 6593 2691 __ 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.
[R] wireframe (reloaded): how to remove the frame around my plot?
Hello, I would like to remove the frame from wireframe, but that one only. I am currently doing trellis.par.set(axis.line,list(col=NA,lty=1,lwd=1)) but this has the flaw that also the tick lines of the axes disapper. I read all the thread with the same titlee of this message appeared sometime ago on this list and tried all the methods there suggested but I am still facing the problem. Is there any possible solution? Thank you for your consideration. Marco --- Marco Chiarandini, Fachgebiet Intellektik, Fachbereich Informatik, Technische Universität Darmstadt, Hochschulstraße 10, D-64289 Darmstadt - Germany, Office: S2/02 Raum E312 Tel: +49.(0)6151.166802 Fax: +49.(0)6151.165326 email: [EMAIL PROTECTED] web page: http://www.intellektik.informatik.tu-darmstadt.de/~machud __ 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
[R] Problems using html and dvips of Hmisc
Dear all, I am experiencing the two following problems trying to produce a comfortable output using the package Hmisc. If I use: html(latex(D,file=here.tex),file=here.html) where D is a data.frame the function html does not write in here.html but only in a temporary file somewhere else in my PC (latex instead writes correctly in the file here.tex. If I use dvips(latex(D),file=here.ps) everything runs smooth but if the data frame is quite large I am not able to center it properly in the output file .ps. The table always stays high on the page going out from the upper side of the paper. I tested this two unwanted behaviours both under linux and under MacOsx and the packages are all up-to-date. Thank you for the consideration. Regards, Marco __ 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
[R] Handling very large integers with factorial and combinat (nCm)
Dear list, perhpas this question is more suitable for R-dev but since I am not really a developer I post it here first. Apparently the following lines do not create any problem in R: library(combinat) r - 20; b - 2; sum( sapply(0:r,function(x) nCm(r,x)^(2*b)) ) 2^64 while in C I obtain an overflow of data even using unsigned long long and with long double I incurr in precision problems. Where can I find information about how R (or the combinat package) handles very large integer numbers? Thank you for consideration, Marco __ 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
[R] Wilcoxon test for mixed design (between-within subjects)
Hallo, is there any extension of the pairwise Wilcoxon test to a dependent samples layout with replicates (or, in other terms, a one-way layout with blocking and replicates)? The Wilcoxon method with matched pairs works for the case of dependent samples with one observation per block, while the Mann-Whitney test works for independent samples, thus one single block and replicated observations. Is there a method which mixes this two cases considering a depedent sample design in which each block has more than one observation? I know it exists a Friedman test for this case but in the Friedman test ranks are constructed considering all subjects jointly, while in Wilcoxon only the pair of subject currently considered are ranked, thus resulting in a more powerful test. If no such method exists in R, I am anyway intersted in possible references. Thank you for the consideration, Regards, Marco __ 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
Re: [R] Wilcoxon test for mixed design (between-within subjects)
Hallo Christoph, There is always the possibility to summarize the replicates and then calculate a common pairwise Wilcoxon test. mmmh, in this case I prefer the Friedman test, I would like not to loose any data. The Friedman test calculates the ranks inside of the blocks, for each block separately and is not considering all subjects jointly. true, but it considers all the treatments together, while Wilcoxon works per pairs of treatments within the blocks. In R implemented is the case with unreplicated blocked data, so this doesn't help you. If no such method exists in R, I am anyway intersted in possible references. Look at: Myles Hollander Douglas A. Wolfe (1999), _Nonparametric statistical methods_. New York: John Wiley Sons. There you find the generalization of the replicated case, but you have to implement it. I already implemented the Friedman test for the replicated case although taken from Conover 1999 _Practical non parametric statistics_. I hope it is the same. My interest in a Wilcoxon procedure is due to the fact that according to Hsu 1996 _Multiple comparisons_ the Friedman test is not recommended for multiple comparisons becasue it has been shown that it is not a confident method (the actual alpha value is greater). Regards, Marco --- Marco Chiarandini, Fachgebiet Intellektik, Fachbereich Informatik, Technische Universität Darmstadt, Hochschulstraße 10, D-64289 Darmstadt - Germany, Office: S2/02 Raum E312 Tel: +49.(0)6151.166802 Fax: +49.(0)6151.165326 email: [EMAIL PROTECTED] web page: http://www.intellektik.informatik.tu-darmstadt.de/~machud __ 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
[R] Empirical cumulative distribution with censored data
Dear list, I would like to plot the empirical cumulative distribution of the time needed by a treatment to attain a certain goal. A number of experiments is run with a strict time limit. In some experiments the goal is attained before the time limit, in other experiments time expires before the goal is attained. The situation is very similar to survivial analysis with censored data. I tryed the function: plot(survfit(Surv(time),data=mydata,conf.int=F)) from the package survival. Nevertheless, what i would like to see is an increase of probability as time increases, and not a decrease of survival probabilty. I tried to play with ecdf(), but dealing with the censored data is quite hard-working in this case. Is there anything for censored data in ecdf like-functions or a way to adapt plot.survfit to my case? Thank you for consideration, Ragards, --- Marco Chiarandini, Fachgebiet Intellektik, Fachbereich Informatik, Technische Universität Darmstadt __ 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
Re: [R] Empirical cumulative distribution with censored data
Peter Dalgaard wrote: Marco Chiarandini [EMAIL PROTECTED] writes: Dear list, I would like to plot the empirical cumulative distribution of the time needed by a treatment to attain a certain goal. A number of experiments is run with a strict time limit. In some experiments the goal is attained before the time limit, in other experiments time expires before the goal is attained. The situation is very similar to survivial analysis with censored data. I tryed the function: plot(survfit(Surv(time),data=mydata,conf.int=F)) from the package survival. Nevertheless, what i would like to see is an increase of probability as time increases, and not a decrease of survival probabilty. I tried to play with ecdf(), but dealing with the censored data is quite hard-working in this case. Is there anything for censored data in ecdf like-functions or a way to adapt plot.survfit to my case? Did you try the fun=event argument? Now yes. It does what I want indeed. Thank you a lot! Marco __ 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
[R] Friedman test for replicated blocked data
Hi, I would need to extend the Friedman test to a replicated design. Currently the function: friedman.test(y, ...) only works for unreplicated designs. I found in Conover 1999 Practical Nonparamteric statistics an extension of the formula to my case. Nevertheless, other sources, like Sheskin 2000 Parametric and Nonparametric statistical Procedures and Daniel 1990 Applied nonparametric statistics give a different formula from that of Conover in the unreplicated case. Since they do not provide the extension of this formula to the replicated case I would like to know if someone could give me a reference where I could find such extension. The formula is indeed very simple: F= z \sqrt{bk(k+1)/6} where z is a quantile from the normal distribution, b the number of blocks and k the number of treatments. Unfortunately, the sources cited do not provide indications from where the formula comes from. Thank you for consideration, Greetings, Marco -- Marco Chiarandini, Technische Universitaet Darmstadt __ [EMAIL PROTECTED] mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
[R] organising the display in Trellis plots
Hallo, I would like to organise at my pleasure the layout of a trellis plot. Currently I have a 3x3 matrix display and 7 plots. Is it possible to choose which specific panels will stay empty? I tried index.perm to arrange the order. Then there is perm.cond which I could not understand if it can serve for my purpose since I cant really find which kind of data it accepts. Thank you for the help. Marco P.S. I attach some code. plot - Dotplot(algo~Cbind(y,lower,upper) | group,data=OUT, pch=3, method=bars,pch.bar=2,lwd=3, cex=0.8, par.strip.text=list(cex=1), panel=function(x,y,...) { print(attr(x,'other')) panel.abline(v=attr(x,'other'),lty=2,col=grey70) panel.Dotplot(x,y,...) }, scales = list(cex=1,rot=c(0,45), x=list(alternating=c(1,1,1,1),limits=c(1,nlevels(OUT$algo)),relation=same), y=list(at=c(1:nlevels(OUT$algo)),labels=levels(OUT$algo),alternating=c(1,1,1,1)) ), main=list(label=), xlab=list(cex=1,label=Average rank), ylab=, aspect=fill,as.table=FALSE, layout=c(3,3) ) print(plot) update(plot,perm.cond=c(3,2,1,4,5,6)) #this does not work update(plot,index.cond=c(3,2,1,4,5,6)) #this works it does not what I wish __ [EMAIL PROTECTED] mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
[R] Re: R package installation
Dear Prof. Johnson, sorry for posting the reply so late. I am already using FINK. But you were right, R was looking for g77 and the gcc libraries under the MacOs distributions. Rather than creating symbolic links I updated FLIBS in /Library/Frameworks/R.framework/Resources/etc/Makeconf so that it searches in the right path /sw/lib/gcc/powerpc-apple-darwin7.2.0/ For g77 I made instead the link you suggest under /usr/bin/. I removed and reinstalled the package and everything run fine, in spite of my versions of libraries and compilers (g77 was already up-date from FINK). Thank you all for the help, Marco Dear Marco, I was given an excerpt with your problem about installing package on a MAC, such as Hmisc. I had the same problems and found a work around. I have not had any trouble loading in source packages since, include Hmisc and Design, acepack and vgam. First, I downloaded and installed the g77 compiler. I use a progam named FINK to find, download and intall g77 (so first I installed FINK then from within FINK I downloaded/installed the g77 compiler.) Do a Google search for FINK, it is easy to find and install. After g77 was installed I had to make a symbolic link so R could find it: ln -s \sw\bin\g77 \usr\bin\g77 (I think I had to make a link to my gcc compiler also) \n -s \sw\bin\gcc \usr\bin\g77 It looks like you already have the g77 compiler from the message. the next mesage you can also remedy by symbolic links. Try ln -s /sw/lib/gcc /usr/local/lib/gcc ln -s /sw/lib/gcc/powerpc-apple-darwin7.5.0 \usr\local\lib\gcc\powerpc-apple-darwin6.8 ln -s /sw/lib/gcc/powerpc-apple-darwin7.5.0/3.4.1 \usr\local\lib\gcc\powerpc-apple-darwin6.8\3.4.2 The first directory path in each of the above may be specific to your configuration for gcc. But this did work for me, and if you find the correct location for gcc/ powerpc-apple-darwinX.Y.Z/U.V.W, you should have no trouble either. Good luck. Sincerely Tim Johnson Adjunct Asst. Professor University of Michigan --- Marco Chiarandini, Fachgebiet Intellektik, Fachbereich Informatik, Technische Universität Darmstadt, Hochschulstraße 10, D-64289 Darmstadt - Germany, Office: S2/02 Raum E317 Tel: +49.(0)6151.166802 Fax: +49.(0)6151.165326 email: [EMAIL PROTECTED] web page: http://www.intellektik.informatik.tu-darmstadt.de/~machud __ [EMAIL PROTECTED] mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
[R] Problems installing packages on MacOS with R 2.00
Dear all, I have a problem installing a package required by Hmisc on MacOS 10.3.5 with R 2.00. g77 -fno-common -g -O2 -c avas.f -o avas.o g77 -fno-common -g -O2 -c rlsmo.f -o rlsmo.o gcc -bundle -flat_namespace -undefined suppress -L/usr/local/lib -o acepack.so ace.o avas.o rlsmo.o -L/usr/local/lib -L/usr/local/lib/gcc/powerpc-apple-darwin6.8/3.4.2 -L/usr/local/lib/gcc/powerpc-apple-darwin6.8/3.4.2/../../.. -lfrtbegin -lg2c -lSystem -lcc_dynamic -framework R ld: warning -L: directory name (/usr/local/lib/gcc/powerpc-apple-darwin6.8/3.4.2) does not exist ld: warning -L: directory name (/usr/local/lib/gcc/powerpc-apple-darwin6.8/3.4.2/../../..) does not exist ld: can't locate file for: -lfrtbegin make: *** [acepack.so] Error 1 ERROR: compilation failed for package 'acepack' I found on the Internet a fix for R 1.8 which suggests to delete the -lfrtbegin library from /Applications/StartR.app/RAqua.app/Contents/etc but this path does not exists anymore on R 2.00. How could I solve the problem. Thank you in advance for the help. Marco - Marco Chiarandini, Fachgebiet Intellektik, Fachbereich Informatik, Technische Universität Darmstadt, Hochschulstraße 10, D-64289 Darmstadt - Germany, Office: S2/02 Raum E317 Tel: +49.(0)6151.166802 Fax: +49.(0)6151.165326 email: [EMAIL PROTECTED] web page: http://www.intellektik.informatik.tu-darmstadt.de/~machud __ [EMAIL PROTECTED] mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Problems installing packages on MacOS with R 2.00
Dear Prof. Ripley, It is R 2.0.0! Your problem is that you do not have g77 installed, or at least, not the same version as was used to compile your version of R. (Please do read the posting guide and tell us where you got R from -- I suspect you did not compile it yourself.) I took R from http://cran.at.r-project.org/. True, I did not compiled it: it was the R 2.0.0 (lastest version) bin package. However, I have g77 version 3.4 (October 2003) installed on MacOS. I deleted both -lfrtbegin and -lg2c from FLIBS in /Library/Frameworks/R.framework/Resources/etc/Makeconf and now Hmisc compiles fine. I could not find a directory /usr/local/lib/gcc/powerpc-apple-darwin6.8/ in my installation. I think this should be in the MacOS X FAQ, but unfortunately the version on CRAN linked from the sidebar and the main FAQ at http://cran.r-project.org/bin/macosx/RMacOSX-FAQ.html is for R 1.9.1, not 2.0.0. Did your installation come with a current version? However, a further problem is that many packages which use Fortran code cannot be compiled for MacOS X as it does not have a shared Fortran run-time library. So I suspect that if you do install g77-3.4.2 you will find that you cannot compile package acepack, and that is why no pre-compiled version of the package is available. --- Marco Chiarandini, Fachgebiet Intellektik, Fachbereich Informatik, Technische Universität Darmstadt, Hochschulstraße 10, D-64289 Darmstadt - Germany, Office: S2/02 Raum E317 Tel: +49.(0)6151.166802 Fax: +49.(0)6151.165326 email: [EMAIL PROTECTED] web page: http://www.intellektik.informatik.tu-darmstadt.de/~machud __ [EMAIL PROTECTED] mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
[R] Problems in installing Rmpi library
dear all, I am trying to install the Rmpi library on a cluster but I obtain the following error, which concerns the dynamic libraries: Rmpi version: 0.4-8 Rmpi is an interface (wrapper) to MPI APIs with interactive R slave functionalities. See `library (help=Rmpi)' for details. Error in dyn.load(x, as.logical(local), as.logical(now)) : unable to load shared library /mypath/.R/library/Rmpi/libs/Rmpi.so: /mypath/.R/library/Rmpi/libs/Rmpi.so: undefined symbol: MPI_Finalize Error in library(Rmpi) : .First.lib failed Error in dyn.unload(x) : dynamic/shared library /mypath/.R/library/Rmpi/libs/Rmpi.so was not loaded Is there anybody who can help me in finding a way to trun around the problem. Thank you for the help, Marco -- Marco Chiarandini, Fachgebiet Intellektik, Fachbereich Informatik, Technische Universität Darmstadt, Hochschulstraße 10, D-64289 Darmstadt - Germany, Office: S2/02 Raum E317 Tel: +49 (0)6151 16-6802 Fax: +49 (0)6151 16-5326 email: [EMAIL PROTECTED] web page: http://www.intellektik.informatik.tu-darmstadt.de/~machud __ [EMAIL PROTECTED] mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Multiple comparisons in a non parametric case
Thanks Rolf and Thomas, It looks to me like what you are doing is trying to judge significance of differences by non-overlap of single-sample confidence intervals. While this is appealing, it's not quite right. Yes, this is what I am trying to do. Apparently, when the replicates are the same for each experimental unit and the experiment is balanced the CI should be the same for all sample-pairs, therefore it is somehow like having single sample CI. I just looked into my copy of Applied Nonparametric Statistics (second ed.) by Wayne W. Daniel (Duxbury, 1990) but that only deals with the situation where there is a single replicate per block-treatment combination (whereas you have 10 reps) and block-treatment interaction is assumed to be non-existent. The problems (or instances of problems) are my blocking factor. But this factor has significant interaction in the ANOVA model. The method that Daniel prescribes in this simple setting seems to be no more than applying the Bonferroni method of multiple comparisons. (Daniel does not say; his book is very much a cook-book.) So you might simply try Bonferroni --- i.e. do all k-choose-2 pairwise comparisons between treatments (using the appropriate 2 sample method for each comparison) doing each comparison at the alpha/k-choose-2 significance level. Where k = the number of treatments = 4 in your case. This method is not going to be super-powerful but it is sometimes surprizing how well Bonferroni stacks up against more ``sophisticated'' methods. I knew about Bonferroni. But I am confused. I have actually two references: Conover Practical Nonparametric statistics (page 371) and Sheskin Handbook and Nonparmetric statistical procedures (page 675). Both these books deal with multiple comparison when the Friedman test would be appropriate. But the formula given are different and the CI I obtain are also different. Sheskin, citing various sources (among them Daniel 1990), uses a formula with the normal distribution z and adjust the alfa value according to Bonferroni (strangely no sample statistic appears in the formula). Conover (which is also a good reference) uses a formula with Student't distribution but does not adjust alfa either in the example he provides where 4 treatments are pairwise compared. The CI I obtain are much smaller if I use the Conover procedure than the Sheskin's. And this happens in spite of the p-adjustment in Sheskin. Smaller CI are for me nicer because I can distinguish better differences But the a factor of 3 between them let me doubt I can really use Conover. Which is your opinion? Thansk again for the help, Ragards, Marco -- Marco Chiarandini, Fachgebiet Intellektik, Fachbereich Informatik, Technische Universität Darmstadt, Hochschulstraße 10, D-64289 Darmstadt - Germany, Office: S2/02 Raum E317 Tel: +49 (0)6151 16-6802 Fax: +49 (0)6151 16-5326 email: [EMAIL PROTECTED] web page: http://www.intellektik.informatik.tu-darmstadt.de/~machud __ [EMAIL PROTECTED] mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
[R] Multiple comparisons in a non parametric case
Dear all, I am conducting a full factorial analysis. I have one factor consisting in algorithms, which I consider my treatments, and another factor made of the problems I want to solve. For each problem I obtain a response variable which is stochastic. I replicate the measure of this response value 10 times. When I apply ANOVA the assumptions do not hold, hence I must rely on non parametric tests. By transforming the response data in ranks, the Friedman test tells me that there is statistical significance in the difference of the sum of ranks of at least one of the treatments. I would like now to produce a plot for the multiple comparisons similar to the Least Significant Difference or the Tukey's Honest Significant Difference used in ANOVA. Since I am in the non parametric case I can not use these methods. Instead, I compare graphically individual treatments by plotting the sum of ranks of each treatment togehter with the 95% confidence interval. To compute the interval I use the Friedman test as suggested by Conover in Practical Nonparametric statistics. I obtain something like this: Treat. A|-+-| Treat. B |-+-| Treat. C |-+-| Treat. D |-+-| The intervals have all the same spread because the number of replications was the same for all experimental units. I would like to know if someone in the list had a similar experience and if what I am doing is correct. In alternative also a reference to another list which could better fit my request is welcome. Thank you for the help, Marco -- Marco Chiarandini, Fachgebiet Intellektik, Fachbereich Informatik, Technische Universität Darmstadt, Hochschulstraße 10, D-64289 Darmstadt - Germany, Office: S2/02 Raum E317 Tel: +49 (0)6151 16-6802 Fax: +49 (0)6151 16-5326 email: [EMAIL PROTECTED] web page: http://www.intellektik.informatik.tu-darmstadt.de/~machud __ [EMAIL PROTECTED] mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
[R] nonparametric two-way structure all-pairwise comparisons
Hello, I am conducting a two-way analysis of variance. The ANOVA assumption are not met, hence I need to use non-parametrical methods. In particular I am interested in all-pairwise comparisons between the levels of one of the two factors. If I transform the data in ranks, can I reuse the TukeyHSD method to produce confidence intervals? If this is not correct, is there any method implemented in R to produce confidence intervals in the non-parametrical context? Thank you for the help! Marco -- Marco Chiarandini, Fachgebiet Intellektik, Fachbereich Informatik, Technische Universität Darmstadt, Hochschulstraße 10, D-64289 Darmstadt - Germany, Office: S2/02 Raum E317 Tel: +49 (0)6151 16-6802 Fax: +49 (0)6151 16-5326 email: [EMAIL PROTECTED] web page: http://www.intellektik.informatik.tu-darmstadt.de/~machud __ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
[R] Trellis plot in multiple display with grid
Dear all, I would like to plot a multipanel display. Each plot should have the curve over time of the solution produced by 10 different algorithms (groups of data). I handle this with: xyplot(quality~time | inst, data=profiles, groups=alg, panel=panel.superpose, type=c(s), as.table=TRUE, scales=list(relation=free,rot=c(0,90),cex=0.8), layout=c(3,3), main=list(label=Run time profile,cex=1.3), xlab=list(label=Normalised time,cex=1.3), ylab=list(label=Number of colours,cex=1.3), par.strip.text=list(cex=0.9), key=list(columns=1,cex=1,space=right, text=list(levels(as.factor(profiles$alg))), #points=Rows(sps,1:9), lines=Rows(spl,1:9) ) ) My problems come when I try to add a grid to each of these plots. I call this: xyplot(quality~time | inst, data=profiles, groups=alg, #panel=panel.superpose, panel=function(x,y) { panel.grid(h = -1, v = -1, lty = 2) panel.xyplot(x,y,type=c(s)) }, as.table=TRUE, scales=list(relation=free,rot=c(0,90),cex=0.8), layout=c(3,4), main=list(label=Run time profile,cex=1.3), xlab=list(label=Normalised time,cex=1.3), ylab=list(label=Number of colours,cex=1.3), par.strip.text=list(cex=0.9), key=list(columns=1,cex=1,corner=c(1,0),x=0.9,y=0.07, text=list(levels(as.factor(profiles$alg))), #points=Rows(sps,1:9), lines=Rows(spl,1:9) ) ) but the result is that I am not anymore able to recognize the different algorithms in each sinlge plot. Is there anyone who can help me? Thank you, Marco Chiarandini -- Marco Chiarandini, Fachgebiet Intellektik, Fachbereich Informatik, Technische Universität Darmstadt, Alexanderstrasse 10, D-64283 Darmstadt - Germany, Office: S1/15 Raum 118 Tel: +49 (0)6151 16-6802 Fax: +49 (0)6151 16-5326 email: [EMAIL PROTECTED] web page: http://www.intellektik.informatik.tu-darmstadt.de/~machud __ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Trellis plot in multiple display with grid
Thank you a lot! It works exactly as I desired. I dare to ask you another detail about Trellis multiple display plots. I would like to plot vertical lines in correspondence of the confidence intervals with the function below in order to make easier the visual comparison. Is this possible? I tried using the same logic proposed in the previous answer but it does not work. Dotplot(as.factor(alg) ~ Cbind(y,lower,upper) | class, data=all,subset=ss, pch=3,method=bars,pch.bar=2, # panel=function(x,y,...) { # panel.abline(v=y) # panel.superpose(x,y,...) # }, scales = list(cex=1, x=list(relation=free), y=list(alternating=c(1,1,1,1), labels=c(levels(all$alg)), at=c(1:nlevels(all$alg, main=list(cex=1.3,label=Multiple comparisons with Tukey's confidence intervals), xlab=list(label=Percentage deviation from best results,cex=1.3,between.columns=5,distance=rep(2,60),size=3), ylab=, aspect=0.6,as.table=TRUE); On Sun, 21 Mar 2004, Deepayan Sarkar wrote: On Sunday 21 March 2004 12:08, Marco Chiarandini wrote: Dear all, I would like to plot a multipanel display. Each plot should have the curve over time of the solution produced by 10 different algorithms (groups of data). I handle this with: xyplot(quality~time | inst, data=profiles, groups=alg, panel=panel.superpose, type=c(s), as.table=TRUE, scales=list(relation=free,rot=c(0,90),cex=0.8), layout=c(3,3), main=list(label=Run time profile,cex=1.3), xlab=list(label=Normalised time,cex=1.3), ylab=list(label=Number of colours,cex=1.3), par.strip.text=list(cex=0.9), key=list(columns=1,cex=1,space=right, text=list(levels(as.factor(profiles$alg))), #points=Rows(sps,1:9), lines=Rows(spl,1:9) ) ) My problems come when I try to add a grid to each of these plots. I call this: xyplot(quality~time | inst, data=profiles, groups=alg, #panel=panel.superpose, panel=function(x,y) { panel.grid(h = -1, v = -1, lty = 2) panel.xyplot(x,y,type=c(s)) }, Instead, try panel=function(x,y,...) { panel.grid(h = -1, v = -1, lty = 2) panel.superpose(x,y,...) }, type = s, as.table=TRUE, scales=list(relation=free,rot=c(0,90),cex=0.8), layout=c(3,4), main=list(label=Run time profile,cex=1.3), xlab=list(label=Normalised time,cex=1.3), ylab=list(label=Number of colours,cex=1.3), par.strip.text=list(cex=0.9), key=list(columns=1,cex=1,corner=c(1,0),x=0.9,y=0.07, text=list(levels(as.factor(profiles$alg))), #points=Rows(sps,1:9), lines=Rows(spl,1:9) ) ) but the result is that I am not anymore able to recognize the different algorithms in each sinlge plot. Is there anyone who can help me? Thank you, Marco Chiarandini -- Marco Chiarandini, Fachgebiet Intellektik, Fachbereich Informatik, Technische Universität Darmstadt, Alexanderstrasse 10, D-64283 Darmstadt - Germany, Office: S1/15 Raum 118 Tel: +49 (0)6151 16-6802 Fax: +49 (0)6151 16-5326 email: [EMAIL PROTECTED] web page: http://www.intellektik.informatik.tu-darmstadt.de/~machud __ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html -- Marco Chiarandini, Fachgebiet Intellektik, Fachbereich Informatik, Technische Universität Darmstadt, Alexanderstrasse 10, D-64283 Darmstadt - Germany, Office: S1/15 Raum 118 Tel: +49 (0)6151 16-6802 Fax: +49 (0)6151 16-5326 email: [EMAIL PROTECTED] web page: http://www.intellektik.informatik.tu-darmstadt.de/~machud __ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Trellis plot in multiple display with grid
This is from Hmisc, right ? I'm not too familiar with it. But I don't see the link with the previous example. What exactly is wrong with this ? Do you want to use a 'groups' argument ? Yes, it is from Hmisc but it requires Lattice and it works with Trellis. I would like that in each of the displays, which are plotted according to class, some vertical lines appear. This vertical lines should be in correspondence of y,lower,upper. I don't know if I need to specifies groups. I tried, but it does not help. I tried using the part that below is commented but I get an error message concering missing groups... Dotplot(as.factor(alg) ~ Cbind(y,lower,upper) | class, data=all,subset=ss, pch=3,method=bars,pch.bar=2, # panel=function(x,y,...) { # panel.abline(v=y) # panel.superpose(x,y,...) # }, scales = list(cex=1, x=list(relation=free), y=list(alternating=c(1,1,1,1), labels=c(levels(all$alg)), at=c(1:nlevels(all$alg, main=list(cex=1.3,label=Multiple comparisons with Tukey's confidence intervals), xlab=list(label=Percentage deviation from best results,cex=1.3,between.columns=5,distance=rep(2,60),size=3), ylab=, aspect=0.6,as.table=TRUE); -- Marco Chiarandini, Fachgebiet Intellektik, Fachbereich Informatik, Technische Universität Darmstadt, Alexanderstrasse 10, D-64283 Darmstadt - Germany, Office: S1/15 Raum 118 Tel: +49 (0)6151 16-6802 Fax: +49 (0)6151 16-5326 email: [EMAIL PROTECTED] web page: http://www.intellektik.informatik.tu-darmstadt.de/~machud __ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html