Dear Ping, I do not know the paper by Calinksi & Harabasz (1974), but the function 'bwss' simply does one-way analysis of variance as explained in any introductory statistics text. I did not actually intend the 'bwss' function to be used for differential expression analyses as you describe. It is there rather as a utility function for the 'anova' function for MAList objects. However you could use it for your data by setting 'x' to be a vector of log-ratios or log-intensities and 'group' to be a vector which takes value '1' for group1, '2' for group2 etc etc. I cannot tell though whether this would be a sensible analysis for your data. Best wishes Gordon At 03:07 PM 10/06/2003, you wrote: >Dear Dr. Smyth, > >I am interested in your R package LIMMA. >I want to use the function BWSS in this package, but >I am not sure the meanings of the arguments of this function > >bwss <- function(x,group). > >Your docu. explains they are: > >x: a numeric vector giving the responses. >group: a vector or factor giving the grouping variable. > >For example, I have a microaray dataset. >gene1, .., gene 10 are grouped to group 1 >gene11,...,gene 20 are grouped to group 2 >gene21,.., gene 31 are grouped to group 3. > >The tumors are t1, t2, .., t10. > >For this dataset, how can I decide these two arguments: x and group???? > > >By the way, the meanings of sums of squares between the group means (bss) >and sums of squares within the groups (wss) are the same as those >in Calinski & Harabasz (1974) CH(K)=(B(k)/(k-1)) / (W(k)/(n-k))????? > >Your kind help is highly appreciated!!!! > >Best regards, > >ping ---------------------------------------------------------------------- ----------------- Dr Gordon K Smyth, Senior Research Scientist, Bioinformatics, Walter and Eliza Hall Institute of Medical Research, 1G Royal Parade, Parkville, Vic 3050, Australia Tel: (03) 9345 2326, Fax (03) 9347 0852, Email: smyth@wehi.edu.au, www: http://www.statsci.org