Why can trying several analyses and reporting only a significant one invalidate the advertised error rate?

• Each analysis retains the advertised rate, so selection leaves the overall rate unchanged
• Selection creates extra chances for a false positive
• Sharing observations makes all analysis results identical, eliminating any selection effect

Which entry belongs in a pre-registration of the primary analysis?

• The outcome, exclusion rule and model
• The model selected later from whichever subgroup has the smallest p-value
• The final hypothesis reconstructed from the pattern observed after collection

Pre-registration prohibits exploratory analyses after collection.

A draft specifies its outcome and sample target but says only 'analyse appropriately'. Which addition resolves the primary-analysis ambiguity?

• Specify model, covariates and missing-data handling
• Specify that several models will be tried and the most significant one called primary
• Specify that the primary model will be chosen after examining the treatment-group results

A team repeatedly applies an ordinary fixed-sample efficacy test and stops at its first significant result. What can happen?

• The stated alpha still bounds the whole procedure because every individual look uses it
• The stopping decision changes power but cannot change the null rejection probability
• The overall false positive rate can exceed alpha

Nonsignificant studies are rarely published. What can a reader conclude about the visible literature?

• It estimates all attempted studies without bias if the published studies report their individual p-values
• It may represent a selected portion of the evidence
• It determines the proportion of false findings without information about unpublished studies

A trial needs safety monitoring and interim efficacy decisions. Which plan addresses the repeated-testing risk?

• Specify safety rules and calibrated sequential efficacy boundaries
• Keep the ordinary efficacy threshold at every look because the maximum sample size was registered
• Adjust only the final efficacy test while allowing earlier stopping at ordinary significance thresholds

A hypothesis was developed after inspecting results but is reported as pre-planned. What correction is needed?

• Call it confirmatory because the hypothesis has a plausible scientific explanation
• Keep the pre-planned label if the reported p-value passes the original threshold
• Report its timing and exploratory status