Practical, careful guidance for planning, measurement, modeling, validation, and reproducible quantitative work.
Search by question, method, model, or research topic.
79 decision guides
A chi-square test works from observed and expected frequencies, not percentages alone. This synthetic 2 × 2 health example shows how to calculate expected counts, diagnose sparse cells, decide when the chi-square approximation is questionable, and use Fisher’s exact test when appropriate.
Read guideA practical framework for determining sample size in clinical and health research by defining the study objective, primary endpoint, meaningful effect or precision target, statistical assumptions, design, allocation, and expected information loss before calculating the required sample.
Read guideA small percentage of missing cells can remove a much larger percentage of participants from a regression. This synthetic case study compares complete-case analysis with multiple imputation and sensitivity analysis—and shows why the conclusion can change.
Read guideLearn how to decide between one-tailed and two-tailed tests based on the research question, how the choice changes rejection regions, p-values, and power, and why the direction must be specified before examining the result.
Read guideAn unusual regression observation may be an outlier, a leverage point, an influential case—or some combination of the three. This synthetic investigation shows how residuals, predictor configuration, Cook’s distance and leave-one-case-out sensitivity analysis answer different questions, and why deleting observations simply to obtain statistical significance is not defensible.
Read guideBootstrap and permutation procedures are both resampling methods, but they answer different inferential questions. This Resource explains when bootstrap methods support estimation and confidence intervals, when permutation tests support null-hypothesis testing, and why the two procedures are not interchangeable.
Read guideLearn how to assess the proportional hazards assumption in Cox regression using graphical checks, Schoenfeld residuals, interval-specific hazard ratios, and formal tests. This Resource explains how to characterize nonproportional hazards and decide whether to use time-varying effects, stratification, or another survival-model framework.
Read guideCompeting events can make ordinary Kaplan–Meier risk estimates misleading when the goal is real-world event probability. This Resource explains how cumulative incidence, cause-specific hazards, and Fine–Gray models answer different scientific questions.
Read guideCronbach’s alpha measures internal-consistency reliability, but it does not by itself prove validity, unidimensionality, item quality, or the correctness of a measurement model. This Resource explains how to interpret alpha alongside item diagnostics, dimensionality, and validity evidence.
Read guideCompare crossover and parallel clinical trial designs by examining within-subject variability, carryover, washout, period effects, dropout, and sample-size efficiency. The key decision is whether repeated treatment periods can support a scientifically valid within-participant comparison.
Read guideA smaller exposure coefficient after adding a mediator does not prove mediation. Learn what causal direct and indirect effects require, including counterfactual definitions, confounding assumptions, temporal ordering, interaction, and the identification challenges of natural effects.
Read guideLearn when a log transformation is methodologically justified in regression and statistical analysis, including decisions based on linearity, changing variance, right-skewness, and multiplicative relationships. The guide also explains what changes after transformation and when rank-based or alternative models may be more appropriate.
Read guideYour design, measurement, data structure, assumptions, and existing output can change the appropriate statistical decision. If the choice is still unclear, discuss the study in context.