AdjPVal

AdjPVal, short for adjusted P-value, is a statistic used to account for multiple hypothesis testing. It is commonly reported in high-throughput analyses such as genomics and transcriptomics, where thousands of individual tests (for example, differential expression of genes) are performed simultaneously.

The adjPVal is derived from a set of raw p-values and is produced by applying a multiple testing correction. The most widely used method is the Benjamini-Hochberg procedure, which controls the false discovery rate (FDR). Other correction methods include Bonferroni, Holm, and Sidak, each with different implications for error control. Depending on the software, the adjusted p-value column may be labeled adj.P.Val or adjPVal.

Interpreting adjPVal involves understanding it as the smallest level of significance at which a given result would be declared significant under the chosen error-rate control. For a common threshold of 0.05, a result with adjPVal < 0.05 is considered significant at the 5% FDR level. It is important to distinguish adjPVal from a raw p-value, which does not account for multiple testing, and from q-values, which are another form of FDR-related measure calculated by different methods.

Limitations include the dependence structure among tests, which can affect the exact guarantees of the correction method. In practice, adjPVal values enable more reliable conclusions in studies with many simultaneous tests by reducing the chance of false positives, while still maintaining reasonable power for discovery. The term is frequently encountered in software outputs, where a column labeled adj.P.Val provides the adjusted significance for each feature.