limma report logFC confidence interval?
2
0
Entering edit mode
@segal-corrinne-4150
Last seen 6.9 years ago
Hi, I too would find it useful to have the CI reported. Thanks, Corrinne -----Original Message----- From: bioconductor-bounces@stat.math.ethz.ch [mailto:bioconductor- bounces@stat.math.ethz.ch] On Behalf Of Gordon K Smyth Sent: 30 October 2010 00:13 To: Timothy Wu Cc: Bioconductor mailing list Subject: [BioC] limma report logFC confidence interval? Sunny's CI is exactly right. CIs could be an option in topTable(), but this the first request for them, so the demand doesn't seem enough for now. Best wishes Gordon > Date: Fri, 29 Oct 2010 00:14:39 -0400 > From: Sunny Srivastava <research.baba at="" gmail.com=""> > To: Timothy Wu <2huggie at gmail.com> > Cc: bioconductor <bioconductor at="" stat.math.ethz.ch=""> > Subject: Re: [BioC] limma report logFC confidence interval? > > Hello Thomas, > I am sure senior members of the list will have more to say, here is my > $0.02. > > logFC is the coefficient of the treatment in your model. Assuming that you > have a model with only one treatment > > log Int_g = b0 + b1 * trt > > b1 = logFC > > The CI of logFC can be found in the same manner as you would do in normal > linear regression, but here instead of usual t(0.975, df) quantile, you > should use the moderated t quantile ie t(0.975, df.residual + df.prior) > > So the 95% CI for logFC will be > > logFC -+ t(0.975, fit3$df.residual + fit3$df.prior) * fit3$stdev.unscaled * > sqrt(fit3$s2.post) > > > Please correct me if I am wrong. > > > Thanks, > S. > > On Thu, Oct 28, 2010 at 7:46 AM, Timothy Wu <2huggie at gmail.com> wrote: > >> HI, >> >> Is there a way to report the CI of logFC from topTable in limma? I googled >> around and it doesn't seem easy to find. I was expecting the option to be >> in >> topTable(). >> >> Thanks, >> >> Timothy ______________________________________________________________________ The information in this email is confidential and intend...{{dropped:9}}
Regression limma Regression limma • 1.0k views
ADD COMMENT
1
Entering edit mode
@gordon-smyth
Last seen 9 hours ago
WEHI, Melbourne, Australia

I have now included an option in topTable() to output 95% confidence intervals. For now, it is available only in the version of limma on the Bioconductor developmental repository.

Note that if you simply want error bars rather than CIs, that standard errors have always been available from limma by

SE <- fit$stdev.unscaled * sqrt(fit$s2.post)

Best wishes
Gordon

> y <- matrix(rnorm(100*3),100,3)
> fit <- eBayes(lmFit(y))
> topTable(fit,confint=TRUE)
    logFC  CI.025 CI.975     t P.Value adj.P.Val     B
79 -1.555 -2.6345 -0.475 -2.82 0.00966     0.493 -4.50
11 -1.293 -2.2625 -0.323 -2.61 0.01555     0.493 -4.51
35 -1.251 -2.2000 -0.302 -2.58 0.01662     0.493 -4.51
97 -1.216 -2.1998 -0.231 -2.42 0.02376     0.493 -4.52
85 -1.171 -2.1252 -0.216 -2.40 0.02467     0.493 -4.52
41 -1.129 -2.1500 -0.108 -2.17 0.04073     0.538 -4.54
82  1.048  0.0944  2.002  2.15 0.04192     0.538 -4.54
49  1.057  0.0893  2.024  2.14 0.04308     0.538 -4.54
10  1.059  0.0402  2.077  2.04 0.05324     0.592 -4.55
60  0.916 -0.0653  1.898  1.83 0.08026     0.803 -4.56
ADD COMMENT
0
Entering edit mode
@richard-friedman-513
Last seen 6.9 years ago
Gordon, I once has a paper returned for lack of CIs and I got out of it by explaining that Limma didn't give them Still it would be helpful if they were available as an option. Often the experimentalists I support want "error bars" (whatever those bars mean) rather than p-values or fdrs. Thanks and best wishes, Rich ----------------------------------------------------------- Richard A. Friedman, PhD Associate Research Scientist, Biomedical Informatics Shared Resource Herbert Irving Comprehensive Cancer Center (HICCC) Lecturer, Department of Biomedical Informatics (DBMI) Educational Coordinator, Center for Computational Biology and Bioinformatics (C2B2)/ National Center for Multiscale Analysis of Genomic Networks (MAGNet) Room 824 Irving Cancer Research Center Columbia University 1130 St. Nicholas Ave New York, NY 10032 (212)851-4765 (voice) friedman at cancercenter.columbia.edu http://cancercenter.columbia.edu/~friedman/ "Did he win the Nobel prize or the Ig Nobel prize for levitating the frog?". Rose Friedman, age 14 On Jan 28, 2011, at 11:13 AM, Segal, Corrinne wrote: > Hi, > > I too would find it useful to have the CI reported. > > Thanks, > > Corrinne > > -----Original Message----- > From: bioconductor-bounces at stat.math.ethz.ch [mailto :bioconductor-bounces at stat.math.ethz.ch > ] On Behalf Of Gordon K Smyth > Sent: 30 October 2010 00:13 > To: Timothy Wu > Cc: Bioconductor mailing list > Subject: [BioC] limma report logFC confidence interval? > > Sunny's CI is exactly right. > > CIs could be an option in topTable(), but this the first request for > them, > so the demand doesn't seem enough for now. > > Best wishes > Gordon > >> Date: Fri, 29 Oct 2010 00:14:39 -0400 >> From: Sunny Srivastava <research.baba at="" gmail.com=""> >> To: Timothy Wu <2huggie at gmail.com> >> Cc: bioconductor <bioconductor at="" stat.math.ethz.ch=""> >> Subject: Re: [BioC] limma report logFC confidence interval? >> >> Hello Thomas, >> I am sure senior members of the list will have more to say, here is >> my >> $0.02. >> >> logFC is the coefficient of the treatment in your model. Assuming >> that you >> have a model with only one treatment >> >> log Int_g = b0 + b1 * trt >> >> b1 = logFC >> >> The CI of logFC can be found in the same manner as you would do in >> normal >> linear regression, but here instead of usual t(0.975, df) quantile, >> you >> should use the moderated t quantile ie t(0.975, df.residual + >> df.prior) >> >> So the 95% CI for logFC will be >> >> logFC -+ t(0.975, fit3$df.residual + fit3$df.prior) * >> fit3$stdev.unscaled * >> sqrt(fit3$s2.post) >> >> >> Please correct me if I am wrong. >> >> >> Thanks, >> S. >> >> On Thu, Oct 28, 2010 at 7:46 AM, Timothy Wu <2huggie at gmail.com> >> wrote: >> >>> HI, >>> >>> Is there a way to report the CI of logFC from topTable in limma? I >>> googled >>> around and it doesn't seem easy to find. I was expecting the >>> option to be >>> in >>> topTable(). >>> >>> Thanks, >>> >>> Timothy > > ______________________________________________________________________ > The information in this email is confidential and intend...{{dropped: > 9}} > > _______________________________________________ > 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
ADD COMMENT
0
Entering edit mode
Hi Gordon, I agree. This would be useful addition as some readers, especially non-statisticians, would like to see the relative errors of the logFCs on plots and tables. Cheers, Fraser -----Original Message----- From: bioconductor-bounces@r-project.org [mailto:bioconductor-bounces at r-project.org] On Behalf Of Richard Friedman Sent: Friday, January 28, 2011 11:21 AM To: Bioconductor mailing list Cc: Gordon Smyth Subject: Re: [BioC] limma report logFC confidence interval? Gordon, I once has a paper returned for lack of CIs and I got out of it by explaining that Limma didn't give them Still it would be helpful if they were available as an option. Often the experimentalists I support want "error bars" (whatever those bars mean) rather than p-values or fdrs. Thanks and best wishes, Rich ----------------------------------------------------------- Richard A. Friedman, PhD Associate Research Scientist, Biomedical Informatics Shared Resource Herbert Irving Comprehensive Cancer Center (HICCC) Lecturer, Department of Biomedical Informatics (DBMI) Educational Coordinator, Center for Computational Biology and Bioinformatics (C2B2)/ National Center for Multiscale Analysis of Genomic Networks (MAGNet) Room 824 Irving Cancer Research Center Columbia University 1130 St. Nicholas Ave New York, NY 10032 (212)851-4765 (voice) friedman at cancercenter.columbia.edu http://cancercenter.columbia.edu/~friedman/ "Did he win the Nobel prize or the Ig Nobel prize for levitating the frog?". Rose Friedman, age 14 On Jan 28, 2011, at 11:13 AM, Segal, Corrinne wrote: > Hi, > > I too would find it useful to have the CI reported. > > Thanks, > > Corrinne > > -----Original Message----- > From: bioconductor-bounces at stat.math.ethz.ch > [mailto:bioconductor-bounces at stat.math.ethz.ch > ] On Behalf Of Gordon K Smyth > Sent: 30 October 2010 00:13 > To: Timothy Wu > Cc: Bioconductor mailing list > Subject: [BioC] limma report logFC confidence interval? > > Sunny's CI is exactly right. > > CIs could be an option in topTable(), but this the first request for > them, so the demand doesn't seem enough for now. > > Best wishes > Gordon > >> Date: Fri, 29 Oct 2010 00:14:39 -0400 >> From: Sunny Srivastava <research.baba at="" gmail.com=""> >> To: Timothy Wu <2huggie at gmail.com> >> Cc: bioconductor <bioconductor at="" stat.math.ethz.ch=""> >> Subject: Re: [BioC] limma report logFC confidence interval? >> >> Hello Thomas, >> I am sure senior members of the list will have more to say, here is >> my $0.02. >> >> logFC is the coefficient of the treatment in your model. Assuming >> that you have a model with only one treatment >> >> log Int_g = b0 + b1 * trt >> >> b1 = logFC >> >> The CI of logFC can be found in the same manner as you would do in >> normal linear regression, but here instead of usual t(0.975, df) >> quantile, you should use the moderated t quantile ie t(0.975, >> df.residual + >> df.prior) >> >> So the 95% CI for logFC will be >> >> logFC -+ t(0.975, fit3$df.residual + fit3$df.prior) * >> fit3$stdev.unscaled * >> sqrt(fit3$s2.post) >> >> >> Please correct me if I am wrong. >> >> >> Thanks, >> S. >> >> On Thu, Oct 28, 2010 at 7:46 AM, Timothy Wu <2huggie at gmail.com> >> wrote: >> >>> HI, >>> >>> Is there a way to report the CI of logFC from topTable in limma? I >>> googled around and it doesn't seem easy to find. I was expecting the >>> option to be in topTable(). >>> >>> Thanks, >>> >>> Timothy > > ______________________________________________________________________ > The information in this email is confidential and intend...{{dropped: > 9}} > > _______________________________________________ > 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
ADD REPLY

Login before adding your answer.

Traffic: 547 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6