Complex Reasoning, Analysis, Research Synthesis & Math
Quality 97/100

Compare results across studies without comparing p-values

Assess agreement between studies by estimating the difference in effects rather than by sorting p-values around a threshold.

A cross-study comparison with an explicit difference estimate, heterogeneity assessment, and a conflict-versus-imprecision verdict.

Template

Role

meta-analytic reviewer

Task

Two or more studies of {{research_question}} report {{results_summary}}. Assess whether their results actually conflict: compute or specify the estimate and interval for the difference between effects, explain why comparing the individual p-values cannot answer the question, and state what combined evidence the set of studies provides.

Context

A press summary claims the studies contradict each other because one crossed the significance threshold and the other did not.

Inputs

  • Per-study point estimates, standard errors, and sample sizes
  • Outcome and exposure definitions per study
  • Population and design differences: {{design_differences}}

Constraints

  • Report the difference estimate with its interval, not a verdict from two p-values
  • Note when identical p-values accompany clearly different effects, and the reverse
  • Flag design or population differences that make pooling inappropriate
  • State whether individually non-significant studies could be jointly informative

Output Format

A comparison table of study estimates, an explicit difference/heterogeneity analysis, and a plain-language verdict on conflict versus agreement.

Quality Criteria

  • Agreement judged on the effect scale, not the p-value scale
  • Heterogeneity analysis specified concretely
  • Pooling caveats explicit
  • Verdict distinguishes 'conflicting' from 'imprecise'
expert
heterogeneity
meta-analysis
replication
statistical-reasoning