Not being a statistician, I was wondering if there were a straightforward answer as to why using a linear model was "better" than a straight mean in limma when estimating the average M value across multiple chips. I believe that the regression analysis is supposed to be less influenced by outliers, is this the chief reason the linear model is implemented? thanks Simon.

I am not quite sure what your model is. When all of the replicates are at the same "level" (i.e. all biological or all technical) and come from one treatment the regression estimate is just the mean. However, once you have several treatments, using the ANOVA model allows you to have a pooled estimate of standard error. And if you have blocks and different types of replicates (e.g. arrays and spots), the ANOVA model allows you to properly account for this in the standard error. Someone else will need to comment on the Bayesian model used in limma, as I am not currently using limma analysis. --Naomi At 09:21 PM 1/11/2004, Simon Melov wrote: >Not being a statistician, I was wondering if there were a straightforward >answer as to why using a linear model was "better" than a straight mean in >limma when estimating the average M value across multiple chips. I believe >that the regression analysis is supposed to be less influenced by >outliers, is this the chief reason the linear model is implemented? > >thanks > >Simon. > >_______________________________________________ >Bioconductor mailing list >Bioconductor@stat.math.ethz.ch >https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor Naomi S. Altman 814-865-3791 (voice) Associate Professor Bioinformatics Consulting Center Dept. of Statistics 814-863-7114 (fax) Penn State University 814-865-1348 (Statistics) University Park, PA 16802-2111