Thanks, David. I am working on a biology-related project where I have a matrix of gene-to-gene interactions, ( interactions * distances ) and depending on the basic R functions, I do get vary different results ... Bogdan On Mon, Sep 10, 2012 at 1:52 PM, David Winsemius <dwinsemius@comcast.net>wrote: > > On Sep 10, 2012, at 1:43 PM, Bogdan Tanasa wrote: > > > Dear all, > > > > please do you have any recommendation about a more advanced function in R > > for lowess/loess regression ? > > the basics lowess() or loess() do not perform as well as I would expect. > > thanks very much, > > > > You should probably look at packages 'mgcv' and 'locfit'. > > (You should also say in what respect you were needing additional > facilities.) > > -- > > David Winsemius, MD > Alameda, CA, USA > > [[alternative HTML version deleted]]

> I am working on a biology-related project > where I have a matrix of gene-to-gene interactions, ( > interactions * distances ) and depending on the basic R > functions, I do get vary different results ... Ermmm.... First, if I read the help pages correctly, the two base loess implementations have different defaults and use quite different calculations, so different results are very much to be expected. Second, the brief description of the poblem made me twitch. What are your coordinates? A matrix of gene interactions _sounds_ like something that one should not put a smoother through; gene ID is usually categorical. To be sensible you'd need some continuous quantitative x- and y- coordinates. S Ellison ******************************************************************* This email and any attachments are confidential. Any use...{{dropped:8}}