Worksheets and research resources · Page 47

Handle Preprints and Later Journal Versions as Linked Evidence

Separate eligibility, version changes, and duplicate participant data when a study appears before and after journal publication.

Connect ATE, ATT and overlap weighting to the population in the research question.

Record who supplied an outcome and avoid silently pooling different perspectives as one measure.

Match the exact instrument, language and administration mode to its current terms of use.

Assign reproducible streams to replicates so parallel workers do not accidentally repeat the same simulated data.

Use explicit logical tests to find contradictory research records without automatically deleting unusual values.

Check exact dependencies in the fitted design rather than assuming a returned coefficient table identifies every parameter.

Check the summary definition and meaningful zero before reporting a relative continuous-outcome comparison.

Check that the reporting window matches the scientific question and remains consistent across assessments.

Use opening dates, ramp-up and mature site rates to estimate calendar recruitment capacity.

Specify episode boundaries, risk periods and terminal events so repeated outcomes represent the intended process.

Check the assignment threshold, local support and sensitivity before interpreting a discontinuity.