[BioC] No $df.residual

Johnson, Franklin Theodore franklin.johnson at email.wsu.edu
Wed May 8 02:32:18 CEST 2013


Awesome, thanks!

Great minds discuss ideas. 
Average minds discuss events.
Small minds discuss people. 
-Eleanor Roosevelt




________________________________________
From: seandavi at gmail.com [seandavi at gmail.com] on behalf of Sean Davis [sdavis2 at mail.nih.gov]
Sent: Tuesday, May 07, 2013 5:30 PM
To: Johnson, Franklin Theodore
Cc: FRANKLIN JOHNSON [guest]; bioconductor at r-project.org; franklin.johnson at wsu.edu
Subject: Re: [BioC] No $df.residual

On Tue, May 7, 2013 at 8:16 PM, Johnson, Franklin Theodore
<franklin.johnson at email.wsu.edu> wrote:
> Dear Sean Davis,
>
> Thanks for the input.
>
> I took the average of the RMA normalized values for each genotype. Then, log2 transformed as below:
>
> The raw data is from NimbleGen array custom-designed for apple genome.
> It was normalized using RMA method. This was downloaded from NCBI/GEO,
> then I took log2(exprs(eset), after reading the papers and making phenoData.txt...Or, I did "log2(matrix(assayData))", then made an eset with this input file.
>
> Anyhow, I will go back and start from the RMA normalized values and replicates.
> However, for 3 reps each for 3 genotypes:

Hi, Franklin.

Take a look at section 8.3 of the limma user guide.  Your experiment
fits that exactly.

> After making an eset,
> This contrast matrix:
> contrastMatrix=c(0,0,1, 0,1,0, 1,0,0, 0,0,2, 0,2,0, 2,0,0, 0,0,3, 0,3,0, 3,0,0)
> should work, right?
> Or, is another format, i.e. -1,.. ; {1,1,1}, a better option for comparision?
> If its 1,1,1, for one genotype, contrast with the other 2 genotypes<- still cannot get $df.residuals correct?
>  Ambrosia Gala Melrose
>> 1        1    0       0
>> 2        1   -1       0
>> 3        1    0      -1
> ---didn't work either---
>
> Is there a way to coerce limma to accept a variable with only 2 levels?

Yes.  Limma accepts factors with only 2 levels--no coercion necessary.
 See section 8.2 of the limma user guide for an example.

> I can write an ANOVA that accepts a factor with just 1 level...it's kinda of odd I can't do this with limma.
> For example, sex is M an F...only 2 levels. If I'm not mistaken, limma require 'more than 2 levels', right?

No, that is not correct.  Limma will accept a factor with only 2 levels.

You will definitely benefit from reading the limma user guide.  Try
following the examples in section 8.2 and 8.3 to get a better sense of
how limma works.

Hope that helps.

Sean

> But, limma is very useful, and efficient.
>
> Hope this helps...I'm here all night!
> Regards,
> Franklin
>
>
> ________________________________________
> From: seandavi at gmail.com [seandavi at gmail.com] on behalf of Sean Davis [sdavis2 at mail.nih.gov]
> Sent: Tuesday, May 07, 2013 3:12 PM
> To: FRANKLIN JOHNSON [guest]
> Cc: bioconductor at r-project.org; franklin.johnson at wsu.edu
> Subject: Re: [BioC] No $df.residual
>
> Hi, Franklin.  See below.
>
> On Tue, May 7, 2013 at 6:04 PM, FRANKLIN JOHNSON [guest]
> <guest at bioconductor.org> wrote:
>>
>> Hello,
>>
>> I want to determine differences between three genotypes.
>> I'm using an exprs(eset).
>> I made:
>> contrastMatrix
>> Ambrosia Gala Melrose
>> 1        1    0       0
>> 2        0    1       0
>> 3        0    0       1
>> However, in fit2()
>> $df.residual
>> [1] 0 0 0 0 0
>> 68660 more elements ...
>>
>> $sigma
>> [1] NA NA NA NA NA
>> 68660 more elements ...
>> I looked back to fit() and had the same output.
>> However, my design matrix should allow me to compute differences. Does it say somewhere reps are necessary, other times seems to suggest need based on hypothesis.
>>
>
> Yes, you need replicates to have residual degrees of freedom and to
> perform a statistical test.
>
>> I already computed the average.
>
> Might be obvious, but the average of what?
>
>> I can also compute the stats for this as well, if needed.
>> Can I create an object of this to use with limma?
>
> Again, can your clarify what data you mean to use with limma?
>
> Sean
>
>> On the other hand,
>> $qr
>>   Ambrosia Gala Melrose
>> 1       -1    0       0
>> 2        0   -1       0
>> 3        0    0       1
>>
>> Also, this contrastMatrix doesn't fit.
>>
>> Regards,
>> Franklin
>>
>>
>>
>>
>>
>>
>>  -- output of sessionInfo():
>>
>> R 2.15.1
>>> objects()
>>  [1] "constrast.matrix"          "contrastNames"             "contrastsMatrix"
>>  [4] "design"                    "eFBestN"                   "eFBestNlog2t"
>>  [7] "eSetlog2_eFBestN"          "exprs"                     "fit"
>> [10] "fit2"                      "geneID"                    "HistogramPlot"
>> [13] "limmaGUIenvironment"       "log2_eFBestN"              "mydesign"
>> [16] "myfit"                     "NEOffsetDefault"           "numParameters"
>> [19] "parameterNames"            "pDataN"                    "pDatarootstocKlog2"
>> [22] "pDatarootstocNlog2"        "pDatascioNlog2"            "phenoDatN"
>> [25] "SampleNames"               "scion.phenoData"           "ss.rootstock.eFBestNlog2t"
>> [28] "ss.rootstock.log2"         "ss.rootstock.log2t"        "ss.scion.eFBestNlog2t"
>> [31] "ss.scion.log2"             "ss.scion.log2t"            "targets"
>> #####################
>> function (package = NULL)
>> {
>>     z <- list()
>>     z$R.version <- R.Version()
>>     z$platform <- z$R.version$platform
>>     if (nzchar(.Platform$r_arch))
>>         z$platform <- paste(z$platform, .Platform$r_arch, sep = "/")
>>     z$platform <- paste(z$platform, " (", 8
>>
>>
>> --
>> 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
>
>
> _______________________________________________
> 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




More information about the Bioconductor mailing list