Dear all, I am starting a statistical analysis for a new experiment and I have some doubts about quality control filtering procedure. I know quality control is essential in order to obtain a "reliable" result at the end of statistical analysis. Usually I discard spots by setting a convenient wt.fun as argument of read.maimages: by this function I assign a weight 0 to each spot with low SNR, or high percentage of saturated pixels, or GenePix "bad" flag, or simlpy if the spot is a ControlSpot (I do not use control spot for normalization process). I read the array quality method of Ritchie et al (2006) implemented in LIMMA. If I do not make a mistake, this method assign an "array weight" to arrays with poor quality (reproducibility) and use it to down-weight their observations. arrayWeights() produces the new array of weights used in lmFit(). I am excited about using this method for my next analysis, but I do not manage to understand how I can put together my personal quality filtering procedure (realized by wt.fun), wich produces an array of weights, and the weights obtained by arrayWeights(). Firstly, is it right putting together the spot weights with array weights? Are the information regarding saturation or low SNR taked into account from arrayWeights() method? Is it essential, in your opinion, using these last two charateristics to discard spots? In my last experiment I discard a lot of spot (about 60%) due to low SNR and I am a bit worried about fitting a model with a so poor number of survived spots.... Or more simply, should I use only arrayWeights() and Spot Type File to eliminate control spots from lmFit? Thank you for any help Erika [[alternative HTML version deleted]]