Practical guides · Page 3
Search and filter all practical guidesInterpret Equivalence Against Prespecified Bounds
Separate evidence of a sufficiently small difference from a failure to reject a zero difference.
Label What Error Bars Actually Represent
Distinguish observed spread, uncertainty in an estimate, and prediction variability in every figure.
Explain How Case Mix Changes External Model Performance
Compare the validation population with the development setting before interpreting a performance difference.
Plan interaction interpretation in a factorial trial
Prespecify factorial comparisons and explain how an interaction changes the treatment contrast.
Write feasibility progression criteria before recruitment
Turn feasibility concerns into prespecified measures, review thresholds and actions for a future trial.
Decide Whether a Finite Population Correction Fits the Sampling Target
Connect the sampling fraction to a defined finite population before reducing a standard error.
Audit Effect Direction Before Drawing a Forest Plot
Align outcome coding, comparators, scales, and favors labels so a forest plot does not reverse its scientific message.
Use Heteroscedasticity-Robust Standard Errors for the Right Problem
Separate a covariance-estimation change from corrections to the mean model, dependence, or confounding.
Find Outcome-Free Time Hidden in Exposure Definitions
Audit future-dependent treatment labels before comparing follow-up outcomes.
Keep Raw Research Exports Separate from Analysis Corrections
Use a write-protected source layer and explicit transformation records so corrections remain reproducible.
Check whether an instrument score crosswalk applies
Verify measure versions, populations and linking uncertainty before converting scores between instruments.
Plan an interrupted time-series analysis
Define the intervention break, trend changes and competing explanations before fitting a time-series model.