Foundations
Read a diagnostic as evidence for a decision, not as a standalone score.
- How to Read DiagnosticsWhat should a statistical diagnostic change about your analysis decision?
Use withNull Hypotheses and P-valuesEffect SizesBootstrap and Permutation Baselines
- Notation and EstimatorsWhat quantity are you trying to learn, what rule estimates it, and what did this particular dataset produce?
Use withMeanVariance and Standard DeviationBootstrap and Permutation Baselines
- Likelihood and Maximum LikelihoodWhich parameter values assign greater joint density or mass to the fixed observations under a declared model, and what does the maximizing value actually establish?
Use withNotation and EstimatorsNull Hypotheses and P-valuesLikelihood-Ratio Test
- Null Hypotheses and P-valuesHow incompatible is the observed statistic with a prespecified null model, and what decision may that incompatibility support?
Use withAugmented Dickey-FullerKPSSKolmogorov-Smirnov Distance
- Likelihood-Ratio TestHow much maximized fit is lost when the null restriction is imposed?
Use withLikelihood and Maximum LikelihoodNull Hypotheses and P-valuesWald Test
- Wald TestHow far is the unrestricted estimate from the null value on its estimated local uncertainty scale?
Use withLikelihood and Maximum LikelihoodNull Hypotheses and P-valuesLikelihood-Ratio Test
- Score TestDoes the likelihood still point away from the null when evaluated at the best null-constrained fit?
Use withLikelihood and Maximum LikelihoodNull Hypotheses and P-valuesLikelihood-Ratio Test
- Likelihood-Based Test ComparisonWhich likelihood-based test matches the fits and evidence available, and why can likelihood-ratio, Wald, and score results separate in finite samples?
Use withLikelihood and Maximum LikelihoodNull Hypotheses and P-valuesLikelihood-Ratio Test
- Effect SizesHow large is the estimated contrast, on which scale, and relative to which decision threshold?
Use withCohen's dWasserstein DistanceKolmogorov-Smirnov Distance
- Bootstrap and Permutation BaselinesHow variable is this estimator across plausible resamples, or how unusual is this statistic after legally breaking the tested relationship?
Use withMutual InformationDistance CorrelationSilhouette Score