[BioC] Designing a model with blocking and other interactions

[Gordon K Smyth]

On Fri, 18 Jul 2014, Eleanor Su wrote:

> Hi Gordon,
>
> I just wanted to elaborate on this email that I got from you back in April.
> This design that you sent me:
>
> design1 <- model.matrix(~Family)
>  design2 <- model.matrix(~mitoHap*Treatment)
>  design <- cbind(design1,design2[,3:4])
>
> Does it account for Family as a blocking factor?

Yes.

> Also, if I increase the number of samples for the analysis (for example 
> 20, instead of 10), does this command below change?
>
> design <- cbind(design1,design2[,3:4])

The design I suggested works for the specific arrangement of samples in 
your experiment.  I cannot vouch that it would be correct for an expanded 
experiment, especially without seeing what treatments are applied to the 
new samples.  It would still work however if the only change is to add 
more families while keeping the same levels for mitoHap and Treatment.

Best wishes
Gordon

> Sorry if these seem like naive questions. I'm just trying to get a better
> understanding of the design matrix. Thanks for your help.
>
> Best,
> Eleanor
>
>
> On Wed, Apr 2, 2014 at 8:25 PM, Gordon K Smyth <smyth at wehi.edu.au> wrote:
>
>> Dear Eleanor,
>>
>>   design1 <- model.matrix(~Family)
>>   design2 <- model.matrix(~mitoHap*Treatment)
>>   design <- cbind(design1,design2[,3:4])
>>
>> Then test for the last coefficient.
>>
>> Best wishes
>> Gordon
>>
>>  Date: Tue, 1 Apr 2014 11:24:52 -0700
>>> From: Eleanor Su <eleanorjinsu at gmail.com>
>>> To: "bioconductor at stat.math.ethz.ch" <bioconductor at stat.math.ethz.ch>
>>> Subject: [BioC] Designing a model with blocking and other interactions
>>>
>>> Hi All,
>>>
>>> I'm trying to set up a model matrix where I can look at the interaction
>>> between Treatment and mitochondrial haplotypes in my paired samples. These
>>> are the preliminary commands that I've set up:
>>>
>>>  rawdata<-read.delim("piRNAtotalcount<10.txt", check.names=FALSE,
>>>>
>>> stringsAsFactors=FALSE)
>>>
>>>> y <- DGEList(counts=rawdata[,2:11], genes=rawdata[,1])
>>>> Family<-factor(c(6,6,9,9,11,11,26,26,28,28))
>>>> Treatment<-factor(c("C","H","C","H","C","H","C","H","C","H"))
>>>> mitoHap<-factor(c("S","S","S","S","S","S","D","D","D","D"))
>>>> data.frame(Sample=colnames(y),Family,Treatment,mitoHap)
>>>>
>>>    Sample Family Treatment mitoHap
>>> 1   6C (S)      6         C       S
>>> 2   6H (S)      6         H       S
>>> 3   9C (S)      9         C       S
>>> 4   9H (S)      9         H       S
>>> 5  11C (S)     11         C       S
>>> 6  11H (S)     11         H       S
>>> 7  26C (D)     26         C       D
>>> 8  26H (D)     26         H       D
>>> 9  28C (D)     28         C       D
>>> 10 28H (D)     28         H       D
>>>
>>>  design<-model.matrix(?)
>>>>
>>>
>>> I have 10 sequencing samples from 5 different families (a treatment and
>>> control sample from each family) and two different types of mitochondrial
>>> haplotypes. How do I set up a design where I can look at the interaction
>>> between the Treatments and mitoHap while still accounting for Family?
>>>
>>> Any help would be greatly appreciated. Thank you for your time.
>>>
>>> Best,
>>> Eleanor

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