Below. -- Bert
On Fri, Oct 7, 2011 at 1:09 PM, Saurav Pathak <sau...@sas.upenn.edu> wrote: > Hi All, > > I am trying to use loess to smooth a 2D image, and also obtain the standard > error for every pixel. I see that the standard error does not make sense. > For example, running the following: > > library(stats) > x <- array(c(1:100), dim=c(100,100)) > y <- t(x) > v <- exp(-((x-50)^2+(y-50)^2)/30^2) > s <- v*0.02 > g_noise <- rnorm(10000, mean = 0, sd = s) > f <- v + g_noise > f.loess <- loess(f ~ x + y, span=0.1, data.frame(x=c(x),y=c(y),f=c(**f))) > f.predict <- predict(f.loess, data = data.frame(x = c(x), y = c(y), f = > c(f)), span = 0.1,se=TRUE) > image(1:100,1:100,matrix(f.**predict$se,nrow=100)) > > I get an image of the standard error that has peaks at regular grid nodes. > Shouldn't I expect to see roughly the same error that I put in (in this > case g_noise)? No. Apparently you do not know the difference between standard error (of a fitted value) and standard deviation. I suggest you consult someone who does. > I notice that the noise peaks move apart for higher span values. > > Thanks for your help! > Saurav > > ______________________________**________________ > R-help@r-project.org mailing list > https://stat.ethz.ch/mailman/**listinfo/r-help<https://stat.ethz.ch/mailman/listinfo/r-help> > PLEASE do read the posting guide http://www.R-project.org/** > posting-guide.html <http://www.R-project.org/posting-guide.html> > and provide commented, minimal, self-contained, reproducible code. > [[alternative HTML version deleted]] ______________________________________________ R-help@r-project.org 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.