Dataset supporting Statistical Departure Vectors: A Domain-Oriented Framework for Detection and Characterization of Heterogeneous Alternatives

SDV represents interpretable departure domains—including location, covariance, distributional shape, and temporal dependence—in a common standardized multivariate space. The resulting departure vector is decomposed into a radial component representing the overall magnitude of departure and a directional component representing its statistical composition.</p><p dir="ltr">The dataset contains results from Monte Carlo simulations across multiple sample sizes, effect sizes, and alternative distributions, including location shifts, scale changes, variance changes, correlation, heavy tails, skewness, mixture distributions, AR(1), MA(1), and change-point processes.

The final simulations used 10,000 null replicates for calibration and 5,000 replicates per alternative condition. Comparisons include SDV-based statistics and established combination procedures such as Fisher, Cauchy, maximum-domain, and minimum-p methods.</p><p dir="ltr">Additional files provide cross-validated classification of departure signatures and analyses of radial and angular behavior across effect sizes. The dataset is intended to support reproducibility and further methodological research on omnibus testing, goodness-of-fit, multivariate model diagnostics, and statistical characterization of model departures.</p>