It is not clear what sort of analysis you are doing, and for example robust/resistant regression is a way of identifying and downweighting outliers in a regression analysis.
Also, multivariate outliers are a very different concept from univariate ones, and the difference may or may not matter depending on the analysis. On Thu, 2 Mar 2006, Robert Lundqvist wrote: > I am sitting with this fairly big material (20 variables, max length of > vectors about 3200 observations and a substantial amount of missing > values). In some cases there are also outliers. Some are obvious, others > are not that clear. > > So far, I have replaced some of the outliers with NA's. However, I would > like to have a good working procedure where outliers where not excluded > permanently but rather temporarily. Some way of "marking" observations > and still keep them seems both preferable and possible. Depends what `keep' means, but in one sense that is what na.action = na.exclude does. > Any suggestions for a good working practice for cases like this? How do > *you* work? Is there any "standard" package to use? > > Robert -- Brian D. Ripley, [EMAIL PROTECTED] 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 272866 (PA) Oxford OX1 3TG, UK Fax: +44 1865 272595 ______________________________________________ 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