betahat = betahat
, sebetahat = betahat / zscore
Last updated: 2017-02-16
Code version: ca57b48
Apply two FDR-controlling procedures, Benjamini–Hochberg 1995 (“BH”) and Benjamini-Yekutieli 2001 (“BY”), and two \(s\) value models, ash
and truncash
(with the threshold \(T = 1.96\)) to the simulated, correlated null data. The data are obtained from 5 vs 5 GTEx/Liver samples and 10K top expressed genes, and \(1000\) independent simulation trials.
Compare the numbers of false discoveries (by definition, all discoveries should be false) obtained by these four methods, using FDR \(\leq 0.05\) and \(s\)-value \(\leq 0.05\) as cutoffs.
ash
and truncash
.\(\hat\beta\) obtained from the limma::lmFit
step of the pipeline, \(\hat z\) obtained from the last step of the pipeline.
library(ashr)
source("../code/truncash.R")
p = read.table("../output/p_null_liver_777.txt")
z = read.table("../output/z_null_liver_777.txt")
betahat = read.table("../output/betahat_null_liver_777.txt")
m = dim(p)[1]
n = dim(p)[2]
fd.bh = fd.by = fd.ash = fd.truncash = c()
for (i in 1:m) {
p_BH = p.adjust(p[i, ], method = "BH")
fd.bh[i] = sum(p_BH <= 0.05)
p_BY = p.adjust(p[i, ], method = "BY")
fd.by[i] = sum(p_BY <= 0.05)
betahat_trial = as.numeric(betahat[i, ])
sebetahat_trial = - betahat_trial / as.numeric(z[i, ])
fit.ash = ashr::ash(betahat_trial, sebetahat_trial, method = "fdr", mixcompdist = "normal")
fd.ash[i] = sum(ashr::get_svalue(fit.ash) <= 0.05)
fit.truncash = truncash(betahat_trial, sebetahat_trial, t = qnorm(0.975))
fd.truncash[i] = sum(get_svalue(fit.truncash) <= 0.05)
}
Simulated under the global null, FWER \(=\) FDR.
fdr.bh = mean(fd.bh >= 1)
fdr.bh
[1] 0.046
Estimated FWER or FDR by BY
fdr.by = mean(fd.by >= 1)
fdr.by
[1] 0.006
Estimated FWER or FDR by ash
fdr.ash = mean(fd.ash >= 1)
fdr.ash
[1] 0.096
Estimated FWER or FDR by truncash
fdr.truncash = mean(fd.truncash >= 1)
fdr.truncash
[1] 0.062
maxcount = max(c(fd.bh, fd.by, fd.ash, fd.truncash))
xlim = c(0, maxcount)
maxfreq = max(c(max(table(fd.bh)), max(table(fd.by)), max(table(fd.ash)), max(table(fd.truncash))))
ylim = c(0, maxfreq)
plot(table(fd.bh), xlab = "Number of False Discoveries / 10K", ylab = "Frequency", main = "Benjamini - Hochberg 1995", xlim = xlim, ylim = ylim)
plot(table(fd.by), xlab = "Number of False Discoveries / 10K", ylab = "Frequency", main = "Benjamini - Yekutieli 2001", xlim = xlim, ylim = ylim)
plot(table(fd.ash), xlab = "Number of False Discoveries / 10K", ylab = "Frequency", main = "ash", xlim = xlim, ylim = ylim)
plot(table(fd.truncash), xlab = "Number of False Discoveries / 10K", ylab = "Frequency", main = "truncash", xlim = xlim, ylim = ylim)
m = length(fd.bh)
fd.ind = (1:m)[!((fd.bh == 0) & (fd.by == 0) & (fd.ash == 0) & (fd.truncash == 0))]
plot(1:length(fd.ind), fd.bh[fd.ind], pch = 4, ylim = xlim, xlab = "Trials with False Discoveries", ylab = "Number of False Discoveries / 10K")
points(1:length(fd.ind), fd.by[fd.ind], pch = 4, col = 2)
points(1:length(fd.ind), fd.ash[fd.ind], pch = 4, col = 3)
points(1:length(fd.ind), fd.truncash[fd.ind], pch = 4, col = 4)
legend("topright", c("BH", "BY", "ash", "truncash"), col = 1:4, pch = 4)
sessionInfo()
R version 3.3.2 (2016-10-31)
Platform: x86_64-apple-darwin13.4.0 (64-bit)
Running under: macOS Sierra 10.12.3
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] SQUAREM_2016.10-1 ashr_2.1.2
loaded via a namespace (and not attached):
[1] Rcpp_0.12.9 knitr_1.15.1 magrittr_1.5
[4] workflowr_0.3.0 MASS_7.3-45 doParallel_1.0.10
[7] pscl_1.4.9 lattice_0.20-34 foreach_1.4.3
[10] stringr_1.1.0 tools_3.3.2 parallel_3.3.2
[13] grid_3.3.2 git2r_0.18.0 htmltools_0.3.5
[16] iterators_1.0.8 yaml_2.1.14 rprojroot_1.2
[19] digest_0.6.9 codetools_0.2-15 evaluate_0.10
[22] rmarkdown_1.3 stringi_1.1.2 backports_1.0.5
[25] truncnorm_1.0-7
This R Markdown site was created with workflowr