Multiple t-tests run a Welch test on every numeric column against a two-class grouping factor, with built-in multiplicity correction and a volcano plot.
Omics-style screening: thousands of variables (genes, proteins, metabolites), one binary phenotype, you want to find the variables that differ. Per-variable Welch t-test is the simplest and most robust default — robust to unequal variances and slightly non-normal data.
With 1000 variables tested at α = 0.05, you'd expect ~50 false positives by chance. Multi-test correction (Benjamini-Hochberg / BY / Bonferroni / Holm) controls either the FDR (BH, BY — more powerful, allows some false discoveries proportionally) or the FWER (Bonferroni, Holm — stricter, no false discoveries). BH is the de facto omics default at q = 0.05.
Volcano plot: x-axis log₂ fold change, y-axis −log₁₀ raw p. Top-right corner = high upregulation + small p; top-left = high downregulation + small p. Standard reporting figure.
Grouping factor (2 classes) + variables to test + optional success-class label + adjustment method.
Per-variable per-class mean + log₂ fold change + raw + adjusted p + volcano plot.
Variables passing both effect-size and adjusted-p thresholds are the candidates worth pursuing in confirmatory experiments.