Dear All: I am using edgeR for RNA-Seq data analysis. According to the paper "Differential expression analysis of multifactor RNA-Seq experiments with respect to biological variation" by McCarthy, DJ et al. (2012), the tagwise dispersion model is recommended. It is a compromise between entirely individual genewise dispersion and the common dispersion models. My question is, is it possible to obtain a "pure genewise" model, i.e. there is NO shrinkage applied towards the common/trended dispersion? I think what I mean is to estimate \phi by maximizing APL_g (\phi) + G_0 * APL_s (\phi) where G_0 is set to zero (no shared part). Which argument in the estimateGLMTagwiseDisp function shall I change? Can I set prior.df = 0 to get a purely genewise model without any shrinkage? In the paper, G_0 = 20/df, but I am not sure if this "df" in the denominator is the prior.df argument in R, or something else. Thank you very much! Best, Gu Sent with Sparrow (http://www.sparrowmailapp.com/?sig) [[alternative HTML version deleted]]

Dear Gu, Yes, you can set prior.df=0 to get genewise dispersion estimates without squeezing, although this not recommended except perhaps as a diagnostic. In the formula G_0=20/df in the paper, df is the residual df and prior.df=20. So prior.df = G_0 * df.residual Best wishes Gordon --------------------------------------------- Professor Gordon K Smyth, Bioinformatics Division, Walter and Eliza Hall Institute of Medical Research, 1G Royal Parade, Parkville, Vic 3052, Australia. http://www.statsci.org/smyth > Date: Sat, 27 Apr 2013 23:46:20 -0700 > From: Gu Mi <neo.migu at="" gmail.com=""> > To: bioconductor at r-project.org > Subject: [BioC] edgeR for entirely individual gene wise dispersion > > Dear All: > > I am using edgeR for RNA-Seq data analysis. According to the paper > "Differential expression analysis of multifactor RNA-Seq experiments > with respect to biological variation" by McCarthy, DJ et al. (2012), the > tagwise dispersion model is recommended. It is a compromise between > entirely individual genewise dispersion and the common dispersion > models. > > My question is, is it possible to obtain a "pure genewise" model, i.e. > there is NO shrinkage applied towards the common/trended dispersion? I > think what I mean is to estimate \phi by maximizing APL_g (\phi) + G_0 * > APL_s (\phi) where G_0 is set to zero (no shared part). Which argument > in the estimateGLMTagwiseDisp function shall I change? Can I set > prior.df = 0 to get a purely genewise model without any shrinkage? In > the paper, G_0 = 20/df, but I am not sure if this "df" in the > denominator is the prior.df argument in R, or something else. > > Thank you very much! > > Best, > Gu > ______________________________________________________________________ The information in this email is confidential and intend...{{dropped:4}}