NA in rlm
On Tue, 19 Mar 2002, Rafael A. Irizarry wrote:
i am using R 1.5 on solaris 2.8 and redhat 7.2 (what i mention below happens for all of these and on R-1.4.1 on solaris 2.8) i am fitting an addtive model (y=roweffect+columneffct+error) to the following data using rlm
This is the rlm in package MASS, I presume? There is no `R 1.5' nor will there ever be. The version of MASS for R 1.5.0 is not yet released (like R 1.5.0).
8.631371 8.219719 12.407817 12.373610 10.558580 10.284926 11.525872 11.504441 12.644998 12.583985 12.772225 12.629980 12.482169 12.271868 12.039199 12.061671 11.892378 11.688070 11.725565 11.655078 11.251226 11.383293 9.199893 7.860003 12.275562 12.240147 13.638981 13.381291 13.205941 13.060880 11.056920 10.707565 if the above matrix is x, when i do the following: m <- dim(x)[1] n <- dim(x)[2] rows<- as.factor(rep(1:m,n)) cols <- as.factor(rep(1:n,rep(m,n))) z <-rlm(as.vector(x)~cols+rows,psi=psi.bisquare) i get an NA in the estimate z$coef[13] this relates to the 12th row effect which appears to be an outlier. is this how one interprets an NA here?
Since you overrode the default psi, you gave zero weight to all observations in row 12. So you have no usable information about row12.
one consequence is that i cant get standard errors through summary.rlm(z). is there a "work around" this?
The work-around is to use a statistical method that fits your problem.
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 _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._