[BioC] limma modeling, paired samples and continuous variable

[Gordon K Smyth]

> Date: Thu, 17 Apr 2014 09:26:33 +0200
> From: Riba Michela <riba.michela at gmail.com>
> To: "James W. MacDonald" <jmacdon at uw.edu>
> Cc: bioconductor at r-project.org
> Subject: Re: [BioC] limma modeling, paired samples and continuous
> 	variable
>
> Hi,
> thanks a lot for your kind answer.
> I have to specify an additional observation:
> the "r"parameter is indeed a numeric variable and also in this situation the result is the same.

Actually it is not possible to get the same message as before if you have 
correctly code r as a numeric variable.

> Would be reasonable to try and model the design as:
> design<- <- model.matrix(~0+r)
> #where "r"is a numeric variable?
>
> for the points about the coefficients I have to reason about

No.

To answer your question "if differential gene expression between two 
classes of disease are correlated with the r status", you probably need a 
Genotype:r iteraction term in your model.

You probably need to show us the whole targets frame for us to help you 
further.  In other words, we need to see:

   data.frame(Genotype,Disease,r)

Best wishes
Gordon

> Thanks a lot
>
> Michela
> Il giorno 15/apr/2014, alle ore 16:23, James W. MacDonald <jmacdon at uw.edu> ha scritto:
>
>> Hi Michela,
>>
>> On 4/15/2014 5:05 AM, michela riba [guest] wrote:
>>> Hi,
>>> I'm sorry for re-posting the message, but I cannot find it in the archive
>>> Thanks a lot for attention
>>>
>>>
>>> Hi,
>>> I would like to model and retrieve differential expression data
>>> regarding an experimental design in which different patients (9) have different disease classes (3 disease classes) and a feature represented with a percentage (0, 0.50, 0.75,1).
>>>     some conditions are replicated 2 or 3 times, regarding the disease condition
>>> Till now I have done an analysis considering Genotype and Disease in the model (as a paired samples analysis)
>>>
>>> design <- model.matrix(~Genotype+Disease)
>>> or
>>> design <- model.matrix(~0+Genotype+Disease)
>>>
>>> now I would like to model also considering
>>> a continuous variable , namely r
>>>
>>> this way: design <- model.matrix(~Genotype+Disease+r)
>>>
>>> to see if differential gene expression between two classes of disease are correlated with the r status
>>>
>>> but till now it is not possible to gain results
>>> Coefficients not estimable: r0,5 r0,75 r1
>>> Warning message:
>>> Partial NA coefficients for 15246 probe(s)
>>
>> This indicates that R is using your 'r' data as factor rather than numeric. I assume that is not what you want? If so you need to ensure that R thinks that 'r' is a numeric vector.
>>
>> If you really are trying to treat 'r' as a factor, then note that you have either an over-specified model (meaning you are trying to estimate more parameters than you have observations), or that three of the coefficients for 'r' are linear combinations of existing coefficients when you already have genotype and disease in the model.
>>
>> Best,
>>
>> Jim
>>
>>
>>>
>>> if I model
>>> design <- model.matrix(~Disease+r)
>>> it goes well, but  it would not consider the different genotypes
>>>
>>> I thank you very much for attention
>>>
>>> Thanks a lot
>>>
>>> Michela
>>>
>>>  -- output of sessionInfo():
>>>
>>> R version 3.0.2 (2013-09-25)
>>> Platform: x86_64-apple-darwin10.8.0 (64-bit)
>>>
>>> locale:
>>> [1] it_IT.UTF-8/it_IT.UTF-8/it_IT.UTF-8/C/it_IT.UTF-8/it_IT.UTF-8
>>>
>>> attached base packages:
>>> [1] stats     graphics  grDevices utils     datasets  methods   base
>>>
>>> other attached packages:
>>> [1] limma_3.18.13
>>>
>>> loaded via a namespace (and not attached):
>>> [1] tools_3.0.2
>>>
>>> --
>>> Sent via the guest posting facility at bioconductor.org.
>>>
>>> _______________________________________________
>>> Bioconductor mailing list
>>> Bioconductor at r-project.org
>>> https://stat.ethz.ch/mailman/listinfo/bioconductor
>>> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor
>>
>> --
>> James W. MacDonald, M.S.
>> Biostatistician
>> University of Washington
>> Environmental and Occupational Health Sciences
>> 4225 Roosevelt Way NE, # 100
>> Seattle WA 98105-6099
>>
>
> Dr. Michela Riba
> Genome Function Unit
> Center for Translational Genomics and Bioinformatics
> San Raffaele Scientific Institute
> Via Olgettina 58
> 20132 Milano
> Italy
>
> lab: +39 02 2643 9114
> skype: mic_mir32
> riba.michela at gmail.com
> riba.michela at hsr.it

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:4}}