I'm having difficulty understanding the "twoD" option of the maNorm function. It seems like the idea is to normalize for two-dimensional spatial artifacts on the cDNA array. Is this correct? Looking at the source code the relevant function would seem to be: loess(z ~ x * y, weights=w, ..... where z = maM values, x = row locations, y = column locations. What I don't understand is how maM values can be normalized relative to the mathematical product of row and column locations (x*y). This would not seem to be a 2D representation of location, e.g. location (7,3) = (3,7). Does the "*" operator have some other meaning here? A second question is whether it is possible to perform 2D loess across an entire array as opposed to within a print-tip group, and whether it is possible to perform normalization based on 2D background measurements instead of foreground measurements?

> I'm having difficulty understanding the "twoD" option of the maNorm > function. It seems like the idea is to normalize for two- dimensional > spatial artifacts on the cDNA array. Is this correct? Looking at > the source code the relevant function would seem to be: > loess(z ~ x * y, weights=w, ..... > where z = maM values, x = row locations, y = column locations. > What I don't understand is how maM values can be normalized relative to > the mathematical product of row and column locations (x*y). This > would not seem to be a 2D representation of location, e.g. location > (7,3) = (3,7). Does the "*" operator have some other meaning here? Yes. In general in the S formula language * represents an interaction term (actually main effects and interaction), not arithmetic multiplication. See the R documentation for formula, and for loess.

Hi Matthew tRMA (http://www.cmis.csiro.au/iap/tRMA/docs/tRMA.pdf) also has a spatial-normalisation function, this one is described in a paper by Wilson et al. in Bioinformatics (june 2003). It is better documented and, in fact, the only one that worked for me since the 2Dloess function in the maNorm package produced M values that varied around -50 instead of 0. regards, _______________________________________________________ Floor Stam Vrije Universiteit Amsterdam Faculty of Earth and Life Sciences Department of Molecular and Cellular Neurobiology De Boelelaan 1085 1081HV Amsterdam The Netherlands Ph: +31-20-4447114 +31-20-5665512 Fax: +31-20-4447112 e-mail: fjstam@bio.vu.nl _______________________________________________________ On 10 May 2004 , at 18:55, Matthew Fero wrote: > I'm having difficulty understanding the "twoD" option of the maNorm > function. It seems like the idea is to normalize for two- dimensional > spatial artifacts on the cDNA array. Is this correct? Looking at > the source code the relevant function would seem to be: > loess(z ~ x * y, weights=w, ..... > where z = maM values, x = row locations, y = column locations. > What I don't understand is how maM values can be normalized relative > to the mathematical product of row and column locations (x*y). This > would not seem to be a 2D representation of location, e.g. location > (7,3) = (3,7). Does the "*" operator have some other meaning here? > > A second question is whether it is possible to perform 2D loess across > an entire array as opposed to within a print-tip group, and whether it > is possible to perform normalization based on 2D background > measurements instead of foreground measurements? > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor >