Hi Sunny, Thanks so much. I am not using LIMMA for microarray data in this case. I like LIMMA because of its capability of modeling and ebays function. It will be a big plus if LIMMA can deal with other type of data. Thanks again and much appreciated. Speaking zero in this case, it is the true observations, not null value. I am not sure how can I ignore them. Steve On Wed, Mar 30, 2011 at 2:01 PM, Sunny Srivastava <research.baba@gmail.com>wrote: > Hi Steve: > > I think LIMMA does the right thing by not having any link/family because it > is meant *only* for log normal intensities. Using it for anything else is > not correct. > > However, as you say you have a lot of 0s -- I would ignore them because you > are trying to extract information where it is not available. You can do > better by something like a zero-inflated modeling. But LIMMA is not meant > for that purpose. > > I must say -- I am surprised you are modelling microarray data and you have > zeros and observation between 0 and 1. > > Thanks, > S. > > > > On Wed, Mar 30, 2011 at 11:55 AM, Steve Shen <sshen@bu.edu> wrote: > >> Hi Sunny, >> >> Thanks so much for your input. Then, how to deal with inf or missing value >> after logit, because the data I am dealing with have lots of zeros in many >> samples. Any opinion? I wish that LIMMA can provide some link/family type of >> arguments just like GLM does. Thanks a lot. >> >> Steve >> >> >> On Wed, Mar 30, 2011 at 6:46 AM, Sunny Srivastava < >> research.baba@gmail.com> wrote: >> >>> Hi Steve: >>> >>> I am sure Prof. Smyth and others will have better insights but here is my >>> $0.02. >>> >>> I would transform the data using the logit transform, that is, >>> >>> x_ij goes to log (x_ij / ( 1 - x_ij) ) ## with boundary value correction >>> when x_ij = 0 or 1 >>> >>> And then use LIMMA. >>> >>> This makes sure that the data you are working with lies in [-Inf, Inf] >>> interval. >>> >>> Thx, >>> S. >>> >>> On Tue, Mar 29, 2011 at 8:44 PM, Steve Shen <sshen@bu.edu> wrote: >>> >>>> Hello list, >>>> >>>> Can LIMMA work with beta distributed interval(0,1) data? Thanks, >>>> >>>> Steve >>>> >>>> [[alternative HTML version deleted]] >>>> >>>> _______________________________________________ >>>> Bioconductor mailing list >>>> Bioconductor@r-project.org >>>> https://stat.ethz.ch/mailman/listinfo/bioconductor >>>> Search the archives: >>>> http://news.gmane.org/gmane.science.biology.informatics.conductor >>>> >>> >>> >> > [[alternative HTML version deleted]]