Can someone please explain the example given in the limma vignette on page 45. It is an example of the classic interaction model. There are two different scenarios that are shown here, one without setting up contrasts, and one with setting up contrasts. My question is specifically regarding the adj. pvalues that are reported. The reported p-values are different for each scenario. Why is that? What is the p-value corresponding to in the first scenario? What is it corresponding to in the second scenario? Here are the results from my data set for scenario 1: ID X.Intercept. density8 treatT density8.treatT AveExpr F P.Value adj.P.Val 8116520 8116520 13.62623 0.06169053 0.061607654 -0.03356050 13.67948 278969.1 2.031311e-37 1.919204e-33 7894098 7894098 13.87349 -0.10169570 0.042710084 -0.01660554 13.83984 276812.7 2.157184e-37 1.919204e-33 8153903 8153903 13.66958 0.05382805 -0.007617839 -0.02515061 13.68640 252543.6 4.391650e-37 1.919204e-33 8038086 8038086 13.65395 0.06105315 0.041262169 -0.06775548 13.68817 248358.7 4.998637e-37 1.919204e-33 8179174 8179174 13.51915 -0.03694281 0.001696369 0.04665425 13.51319 242354.9 6.042135e-37 1.919204e-33 When i set up the contrasts as shown in the example, and pull out info. for the first probeset id in the list above(8116520), the p-values are different: ID TvsUinlowDensity TvsUinhighDensity Diff AveExpr F P.Value adj.P.Val 8116520 8116520 0.06160765 0.02804716 -0.0335605 13.67948 1.707745 0.2136743 0.3896945 I also have two conditions density and treatment. Any insight/clarification will be appreciated. Thanks! -- output of sessionInfo(): R version 2.15.1 (2012-06-22) Platform: x86_64-apple-darwin9.8.0/x86_64 (64-bit) locale: [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8 attached base packages: [1] stats graphics grDevices utils datasets methods base other attached packages: [1] limma_3.14.1 hugene10stv1cdf_2.11.0 AnnotationDbi_1.20.3 affy_1.36.0 Biobase_2.18.0 [6] BiocGenerics_0.4.0 loaded via a namespace (and not attached): [1] affyio_1.26.0 BiocInstaller_1.8.3 DBI_0.2-5 IRanges_1.16.4 parallel_2.15.1 [6] preprocessCore_1.20.0 RSQLite_0.11.2 stats4_2.15.1 tools_2.15.1 zlibbioc_1.4.0 -- Sent via the guest posting facility at bioconductor.org.

On 3/18/13 12:27 PM, limmauser [guest] wrote: > Can someone please explain the example given in the limma vignette on page 45. It is an example of the classic interaction model. There are two different scenarios that are shown here, one without setting up contrasts, and one with setting up contrasts. > > My question is specifically regarding the adj. pvalues that are reported. The reported p-values are different for each scenario. Why is that? What is the p-value corresponding to in the first scenario? What is it corresponding to in the second scenario? When you generate a topTable and you don't specify a contrast, then you get an F-test in which you are testing that any of the coefficients not equal to zero. This doesn't make any sense in the first case, where you have an intercept, because the intercept is estimating the mean expression of one sample type. You don't really care if the mean expression value is equal to zero or not; instead you are interested in knowing if the difference between two sample types is equal to zero. In other words, microarray data are not meaningful except in the context of a comparison between samples. If you do something like topTable(fitfromscenario1, coef=2:4) You should get the same results as scenario 2. Best, Jim > > Here are the results from my data set for scenario 1: > ID X.Intercept. density8 treatT density8.treatT AveExpr F P.Value adj.P.Val > 8116520 8116520 13.62623 0.06169053 0.061607654 -0.03356050 13.67948 278969.1 2.031311e-37 1.919204e-33 > 7894098 7894098 13.87349 -0.10169570 0.042710084 -0.01660554 13.83984 276812.7 2.157184e-37 1.919204e-33 > 8153903 8153903 13.66958 0.05382805 -0.007617839 -0.02515061 13.68640 252543.6 4.391650e-37 1.919204e-33 > 8038086 8038086 13.65395 0.06105315 0.041262169 -0.06775548 13.68817 248358.7 4.998637e-37 1.919204e-33 > 8179174 8179174 13.51915 -0.03694281 0.001696369 0.04665425 13.51319 242354.9 6.042135e-37 1.919204e-33 > > When i set up the contrasts as shown in the example, and pull out info. for the first probeset id in the list above(8116520), the p-values are different: > > ID TvsUinlowDensity TvsUinhighDensity Diff AveExpr F P.Value adj.P.Val > 8116520 8116520 0.06160765 0.02804716 -0.0335605 13.67948 1.707745 0.2136743 0.3896945 > > I also have two conditions density and treatment. Any insight/clarification will be appreciated. > > Thanks! > > > -- output of sessionInfo(): > > R version 2.15.1 (2012-06-22) > Platform: x86_64-apple-darwin9.8.0/x86_64 (64-bit) > > locale: > [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8 > > attached base packages: > [1] stats graphics grDevices utils datasets methods base > > other attached packages: > [1] limma_3.14.1 hugene10stv1cdf_2.11.0 AnnotationDbi_1.20.3 affy_1.36.0 Biobase_2.18.0 > [6] BiocGenerics_0.4.0 > > loaded via a namespace (and not attached): > [1] affyio_1.26.0 BiocInstaller_1.8.3 DBI_0.2-5 IRanges_1.16.4 parallel_2.15.1 > [6] preprocessCore_1.20.0 RSQLite_0.11.2 stats4_2.15.1 tools_2.15.1 zlibbioc_1.4.0 > > > -- > 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