[BioC] LIMMA paired T-test

[Gordon K Smyth]

Dear Som,

I certainly do not recommend Welch's t-test.

Your limma analysis is already full adjusting for patient variability, and 
Welch's test has nothing to do with patient to patient variability anyway.

Best wishes
Gordon

---------------------------------------------
Professor Gordon K Smyth,
Bioinformatics Division,
Walter and Eliza Hall Institute of Medical Research,
1G Royal Parade, Parkville, Vic 3052, Australia.
Tel: (03) 9345 2326, Fax (03) 9347 0852,
http://www.statsci.org/smyth

On Fri, 6 Jul 2012, somnath bandyopadhyay wrote:

>
>
>
> Hi
> Gordon,
>
> Thanks for your suggestion. That helped a lot!
>
>
>
> I had one more question: if the patient to patient variability is too large,
> would you recommend doing a Welch's t-test? Is there a way to do it in limma
> using the same linear model (~patient + dis + dis:tx)?
>
>
>
> Thanks,
>
> Som.
>
>
>
>> Date: Wed, 4 Jul 2012 10:27:18 +1000
>> From: smyth at wehi.EDU.AU
>> To: genome1976 at hotmail.com
>> CC: bioconductor at r-project.org; maintainer at bioconductor.org
>> Subject: Re: LIMMA paired T-test
>>
>> Your design matrix is not sufficient to answer questions 2 and 3.  Your
>> questions presume an interaction between treatment and disease, i.e.,
>> distinct effects for treatment for disease and healthy, whereas your model
>> formula assumes no interaction.
>>
>> You need:
>>
>>    design <- model.matrix(~patient + dis + dis:tx)
>>
>> Then last two coefficients answer questions 2 and 3.
>>
>> Gordon
>>
>> ---------------------------------------------
>> Professor Gordon K Smyth,
>> Bioinformatics Division,
>> Walter and Eliza Hall Institute of Medical Research,
>> 1G Royal Parade, Parkville, Vic 3052, Australia.
>> http://www.wehi.edu.au
>> http://www.statsci.org/smyth
>>
>> On Tue, 3 Jul 2012, somnath bandyopadhyay wrote:
>>
>>>
>>> Hi Gordon and LIMMA users,
>>>
>>> I am sure this question has been answered before and I tried looking into the archives for some answer but did n't have any success there.
>>>
>>> My experimental design has diseased and healthy volunteers blood treated with a drug. I have gene expression data for both before and after treatment. So, I have disease, treatment and patient_ID (before vs. after treatment) as covariates. What I am interested in are as follows:
>>>
>>> 1. What genes change in untreated disease vs. untreated healthy volunteers?
>>> 2. What genes change in treated disease vs. untreated disease blood samples?
>>> 3. What genes change in treated healthy volunteers vs. untreated healthy volunteers blood samples?
>>>
>>> Design of the experiment:
>>> design <- model.matrix(~ dis + tx + patient)
>>>
>>> Based on the above design I am able to answer question 1. I was
>>> wondering how I would answer question 2 and 3 in a paired T -test (to
>>> account for before vs. after treatment). Do I need to do some contrasts
>>> because I have been trying to work off the lmfit.
>>>
>>> Any help would be greatly apreciated.
>>>
>>> Thanks,
>>> Som.
>>>
>>>
>>>
>>>
>>>
>>
>> ______________________________________________________________________
>> The information in this email is confidential and intended solely for the addressee.
>> You must not disclose, forward, print or use it without the permission of the sender.
>> ______________________________________________________________________
>

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