[BioC] Help: LIMMA-decideTests and lfc option

Gordon K Smyth smyth at wehi.EDU.AU
Mon Jan 26 02:02:23 CET 2009

Dear Sherosha,

If you use

   topTable(fit, coef=c(1,2,3), lfc=1)

then the genes with be ranked by F-statistic, and any probe satisfying lfc 
for any one contrast will be considered.

Regarding decideTests() and method="separate", it doesn't make sense to me 
to use a multiple-contrast cutoff unless you also use a multi-contrast 
test statistic.

Best wishes

> Date: Fri, 23 Jan 2009 17:40:07 -0500
> From: Sherosha Raj <sherosha at gmail.com>
> Subject: Re: [BioC] Help: LIMMA-decideTests and lfc option
> To: James MacDonald <jmacdon at med.umich.edu>,
> 	bioconductor at stat.math.ethz.ch
> Message-ID:
> 	<6cdc8dd00901231440w44b651f4l720d6af746771dfb at mail.gmail.com>
> Content-Type: text/plain
> Hi Jim
> Thank you for the clarification. Is it possible to have a conditional
> selection based on the lfc in decideTests?
> Eg. if I had three contrasts and there is a probeset with lfc that satisfies
> my cutoffs in just ONE contrast but not in the others, and I would still
> like it to be considered?
> Thank you again.
> Regards,
> Sherosha
> 2009/1/23 James MacDonald <jmacdon at med.umich.edu>
>> Hi Sherosha,
>> The lfc argument applies to all the contrasts you have made individually.
>> So for instance if you have three contrasts and for one probeset they were
>> (-1.5, 0.3, and 1.1) and you set lfc = 1, then the first and third would
>> pass that filter and the second would not.
>> Best,
>> Jim
>> James W. MacDonald, M.S.
>> Biostatistician
>> Hildebrandt Lab
>> 8220D MSRB III
>> 1150 W. Medical Center Drive
>> Ann Arbor MI 48109-0646
>> 734-936-8662
>>>>> Sherosha Raj <sherosha at gmail.com> 01/22/09 2:28 PM >>>
>> Hello everyone
>> I am unsure how to interpret the lfc option in the multiple testing
>> algorithm decideTests:
>> decideTests(object,method="separate",adjust.method="BH",p.value=0.05,lfc=0)
>> The explanation given is:
>> "If lfc>0 then contrasts are judged significant only when the
>> log2-fold change is at least this large in absolute value. For
>> example, one might choose lfc=log2(1.5) to restrict to 50% changes or
>> lfc=1 for 2-fold changes. In this case, contrasts must satisfy both
>> the p-value and the fold-change cutoff to be judged significant."
>> What is meant exactly? Is it within a particular contrast that the
>> log2FC>0 for a particular probeset? Or is it comparing across all the
>> contrasts specified?
>> If some one could offer a clarification  I would be most grateful.
>> Thank you.
>> --
>> Regards,
>> Sherosha

More information about the Bioconductor mailing list