A search of the archives did not reveal an answer: For basic tests of association, where one has no a priori knowledge of the form of the relation or of the distributions of the variables, rank correlation seems like a good start. Why is cor.test() with Kendall and Spearman options relegated to the ctest package, rather than in the base package? Does this suggest that the developers consider other tests of association more generally useful? Thanks, Hank Martin Henry H. Stevens, Assistant Professor 338 Pearson Hall Botany Department Miami University Oxford, OH 45056 Tel: (513) 529-4206 FAX: (513) 529-4243 http://www.muohio.edu/~botcwis/bot/henry.html -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-help-request at stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
"Martin Henry H. Stevens" <HStevens at muohio.edu> writes:> A search of the archives did not reveal an answer: > For basic tests of association, where one has no a priori knowledge of > the form of the relation or of the distributions of the variables, > rank correlation seems like a good start. Why is cor.test() with > Kendall and Spearman options relegated to the ctest package, rather > than in the base package? Does this suggest that the developers > consider other tests of association more generally useful?No. *All* "classical tests" are in ctest. Also t.test for instance. Arguably other stuff (e.g. lm/glm) ought to be moved away from base and automatically loaded upon startup in the same way that ctest is, leaving a "lean, mean, and clean" core, that could be used efficiently in e.g. shell scripts. (R has a start-up time of about a second on current hardware - give or take a few powers of 2, which is generally tolerable unless running in a loop.) -- O__ ---- Peter Dalgaard Blegdamsvej 3 c/ /'_ --- Dept. of Biostatistics 2200 Cph. N (*) \(*) -- University of Copenhagen Denmark Ph: (+45) 35327918 ~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk) FAX: (+45) 35327907 -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-help-request at stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
On Thu, 25 Apr 2002, Martin Henry H. Stevens wrote:> A search of the archives did not reveal an answer: > For basic tests of association, where one has no a priori knowledge of the > form of the relation or of the distributions of the variables, rank > correlation seems like a good start. Why is cor.test() with Kendall and > Spearman options relegated to the ctest package, rather than in the base > package? Does this suggest that the developers consider other tests of > association more generally useful?It's *not* relegated. t.test is there too! There is no sense in which material in the standard or recommended packages is not considered `generally useful'. However, R does keep all its code in memory, and the long-term aim has been to strip the base package down to the bare essentials (e.g. removing lm). Many R users never use lm nor cor.test, and I suspect the vast majority of R sessions do not use either. If we had all the standard material in base, R would run slower and need more memory (and the latter has been an issue until recently, with 16Mb teaching labs). -- Brian D. Ripley, ripley at stats.ox.ac.uk Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272860 (secr) Oxford OX1 3TG, UK Fax: +44 1865 272595 -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-help-request at stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
On 04/25/02 05:20, Martin Henry H. Stevens wrote:>A search of the archives did not reveal an answer: >For basic tests of association, where one has no a priori knowledge of the >form of the relation or of the distributions of the variables, rank >correlation seems like a good start. Why is cor.test() with Kendall and >Spearman options relegated to the ctest package, rather than in the base >package? Does this suggest that the developers consider other tests of >association more generally useful? >Thanks, >HankI think I remember a version of R in which the ctest package had to be loaded with library(), but it was a long time ago. Everything in the ctest package is now simply available when you start R. I suspect that it is separate from "base" for historical reasons. It isn't clear to me whether change is needed, and, if so, whether the base package should become even larger or whether it should be broken up into yet more parts. As things are now, even on a fast computer loading from its own disk, the html listing of the base package takes a bit of time to load. My own opinion is that this is a low-priority issue. As for what to do when you don't know what your data look like, my own recommendation is not to do anything in cor.test(), but, use plot(), and then figure out what to do next, e.g., fix errors, apply a tranform, or do a test. Sometimes, of course, you know in advance that tau is the appropriate test. -- Jonathan Baron, Professor of Psychology, University of Pennsylvania Home page: http://www.sas.upenn.edu/~baron -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-help-request at stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._