fdr.adjust {LPE} | R Documentation |
Based on the type of adjustment, eg: resampling, BH, BY, etc, calls appropriate functions for fdr adjustment
fdr.adjust(lpe.result,adjp="resamp",target.fdr=c(10^-3 ,seq(0.01,0.10,0.01), 0.15, 0.20, 0.50),iterations=5,ALL=FALSE )
lpe.result |
Data frame obtained from calling lpe function |
adjp |
Type of adjustment procedure. Can be "resamp", "BH", "BY", "Bonferroni" or "mix.all" |
target.fdr |
Desired FDR level (used only for resampling based adjustment) |
iterations |
Number of iterations for stable z-critical. |
ALL |
If TRUE, the FDR corresponding to all the z-statistics, i.e. for every gene intensity is given. |
Returns the output similar to lpe function, including adjusted FDR. BH and BY give Benjamini-Hochberg and Benjamini-Yekutieli adjusted FDRs (adopted from multtest procedure), Bonferroni adjusted p-values and "mix.all" gives SAM-like FDR adjustment. For further details on the comparisons of each of these methods, please see the reference paper (Rank-invariant resampling...) mentioned below. Users are encouraged to use FDR instead of Bonferrni adjsusted p-value as initial cutoffs while selecting the significant genes. Bonferroni adjusted p-values are provided under Bonferroni method here just for the sake of completion for the users who want it.
Nitin Jainnitin.jain@pfizer.com
J.K. Lee and M.O.Connell(2003). An S-Plus library for the analysis of differential expression. In The Analysis of Gene Expression Data: Methods and Software. Edited by G. Parmigiani, ES Garrett, RA Irizarry ad SL Zegar. Springer, NewYork.
Jain et. al. (2003) Local pooled error test for identifying differentially expressed genes with a small number of replicated microarrays, Bioinformatics, 1945-1951.
Jain et. al. (2005) Rank-invariant resampling based estimation of false discovery rate for analysis of small sample microarray data, BMC Bioinformatics, Vol 6, 187.
# Loading the library and the data library(LPE) data(Ley) dim(Ley) # Gives 12488*7 # First column is ID. Ley[,2:7] <- preprocess(Ley[,2:7],data.type="MAS5") # Subsetting the data subset.Ley <- Ley[1:1000,] # Finding the baseline distribution of condition 1 and 2. var.1 <- baseOlig.error(subset.Ley[,2:4], q=0.01) var.2 <- baseOlig.error(subset.Ley[,5:7], q=0.01) # Applying LPE lpe.result <- lpe(subset.Ley[,2:4],subset.Ley[,5:7], var.1, var.2, probe.set.name=subset.Ley[,1]) final.result <- fdr.adjust(lpe.result, adjp="resamp", target.fdr=c(0.01,0.05), iterations=1) final.result