qvalue

Dependencies & Environment

Package-intrinsic requirements from the Bioconductor landing page, reproduce in any R environment.

• Version: 2.44.0 · Bioconductor: 3.23 · R: ≥ 4.6
• Imports: ggplot2, reshape2
• Install: BiocManager::install("qvalue")

When to Use

• False Discovery Rate Estimation: When analyzing thousands of features (e.g., differential gene expression) to estimate q-values and control the false discovery rate using qvalue.
• Estimating True Nulls: When you need to estimate the overall proportion of true null hypotheses ($\pi_0$) from a distribution of p-values using pi0est.
• Local FDR Calculation: When you want to calculate the empirical Bayesian posterior probability that a null hypothesis is true, conditional on the observed p-value, using lfdr.
• Empirical P-value Calculation: When you have observed test-statistics and simulated/resampled null statistics and need to compute p-values using empPvals.

When NOT to Use

• Dependent Test Statistics: When tests are strongly dependent or have uncorrected batch effects (indicated by a U-shaped p-value histogram), use the sva package first to correct the data.
• Small-scale hypothesis testing: For experiments with very few tests, use standard Benjamini-Hochberg adjustment because the estimation of $\pi_0$ is highly unstable with small sample sizes.

Data Requirements

• A numeric vector of p-values strictly bounded between 0 and 1, OR a vector of observed statistics and a matrix of empirical null statistics.
• P-values must follow a Uniform(0,1) distribution under the null hypothesis (the right tail of the p-value histogram should be fairly flat).