[BioC] Fwd: Re: implementing limma with several sistematic effects

[Naomi Altman]

>To: Pedro López Romero <plopez at cnic.es>
>From: Naomi Altman <naomi at stat.psu.edu>
>Subject: Re: [BioC] implementing limma with several sistematic effects
>Cc:
>Bcc:
>X-Eudora-Signature: <work>
>Date: Fri, 10 Mar 2006 16:21:54 -0500
>
>I would say that the problem is that the 6 cell 
>lines are nested in the 3 sib pairs.  The 
>simplest way to handle this is just to leave out the sib effect.
>If you want to estimate the sib effect, you need 
>to code the cell lines with only 2 levels (sib A 
>and sib B in each pair) and include the sib*cell 
>line interaction.  The cell line "main effect" 
>and interaction do not really have a meaning in 
>this context, but the two effects together are 
>the cell line effect you tried to estimate in your set-up below.
>
>--Naomi
>
>
>At 03:17 PM 3/10/2006, you wrote:
>>Dear list,
>>
>>I am triying to use limma to compare 6 different treatments.- According to
>>how the data have been generated, appart from the treatment effect, there
>>are two other clear effects 1) due to the cell lines where the treatments
>>have been applied and 2) due to the genetic relationship between the animals
>>where the cells were extracted.- Well, I would like to make a comparison
>>between treatments taking into account the variability that these tow
>>additional sistematic effects introduce in the data.-  The model equation
>>would be:
>>
>>Y = treatment + cell +  sib + error .-
>>
>>I am constructing the design matrices manually as follows:
>>
>>sibEFF=factor(c("mr1","mr1","mr1","mr2","mr2","mr2","mr3","mr3","mr3",
>>"mr1","mr1","mr1","mr2","mr2","mr2","mr3","mr3","mr3"),
>>levels=c("mr1","mr2","mr3"))
>>
>>claEFF=factor(c("c1","c1","c1","c2","c2","c2","c3","c3","c3","c4","c4","c4",
>>"c5","c5","c5","c6","c6","c6"), levels=c("c1","c2","c3","c4","c5","c6"))
>>
>>
>>ttoEFF=factor(c("t1","t2","t3","t1","t2","t3","t1","t2","t3","t4",
>>"t5","t6","t4","t5","t6","t4","t5","t6"),
>>levels=c("t1","t2","t3","t4","t5","t6"))
>>
>>design=model.matrix(~   - 1 + ttoEFF + claEFF + sibEFF)
>>CM= makeContrasts(t1-t2,levels=ttoEFF)
>>fit=lmFit(eset,design)
>>
>>I do not know how the parameters of the model are estimated. I guess that
>>with this model equation,  the X’X matrix is not singular and some
>>coefficients of the cell and sib effects can not be estimated.-
>>
>> >From here I get an error message when I apply fit2=contrasts.fit(fit,CM). I
>>have to be doing something wrong but I do not know what it is.
>>
>>I was thinking to fit first the linear model
>>
>>Y = cell +  sib + error
>>and from here I think I could use limma with the estimated error terms to
>>get the DE genes due to the treatment effect. Could be this a good strategy?
>>
>>
>>I would appreciate very much if someone could give me any advice.
>>
>>Other think that I would like to know is if it s possible to check the
>>quality of the limma models (some residual analysis, QQ-plots, BIC ,
).
>>
>>Thanks for any suggestion.-
>>
>>Pedro.-
>>
>>
>>
>>
>>
>>
>>
>>         [[alternative HTML version deleted]]
>>
>>_______________________________________________
>>Bioconductor mailing list
>>Bioconductor at stat.math.ethz.ch
>>https://stat.ethz.ch/mailman/listinfo/bioconductor
>
>Naomi S. Altman                                814-865-3791 (voice)
>Associate Professor
>Dept. of Statistics                              814-863-7114 (fax)
>Penn State University                         814-865-1348 (Statistics)
>University Park, PA 16802-2111

Naomi S. Altman                                814-865-3791 (voice)
Associate Professor
Dept. of Statistics                              814-863-7114 (fax)
Penn State University                         814-865-1348 (Statistics)
University Park, PA 16802-2111