Practical guides · Page 9
Search and filter all practical guidesCompare Simulation Bias and RMSE Without Treating Them as Interchangeable
Measure systematic error and total squared error against the scenario-specific truth.
Compare Reported Standard Errors with Empirical Simulation Variation
Use between-replicate spread to assess whether model-based uncertainty is calibrated.
Evaluate Interval Coverage and Width Together
Check how often simulated intervals contain the truth while preserving failed intervals and interval lengths.
Count Method Failures as Simulation Results
Track nonconvergence and invalid outputs without letting successful subsets make a method look artificially reliable.
Build a Simulation Scenario Grid Before Running Methods
Separate scientific conditions, performance measures, and computational replication in a traceable design.
Document Spline Knots and Reuse the Fitted Basis
Preserve spline construction so predictions use the same transformation as the fitted model.
Audit Stabilized Inverse-Probability Weights From Numerator to Outcome Model
Check probability definitions, weight distributions and the assumptions retained after stabilization.
Separate calendar time from stepped-wedge exposure
Audit crossover timing, calendar periods and correlation before analysing a stepped-wedge trial.
Prepare a Precise Request for Missing Study Information
Specify the missing quantity, its intended use, and a traceable response record before contacting study investigators.
Check Design Degrees of Freedom Before Using a Large-Sample Interval
Relate survey uncertainty to independent sampling information rather than the number of records alone.
Estimate a Survey Subpopulation Without Losing Its Design
Define the target domain while retaining the sampling information needed for uncertainty.
Distinguish Empty Survey Cells From Observed Zero Outcomes
Keep absent information visible when reporting subgroup estimates and tables.