Effect Size Interpretation Chain

validated scientific chain v1.0.0 cc-by-sa

Block ID: 27316191-b686-4c99-aaa0-3023585684b6

Community-contributed block. PromptDNA makes no guarantee of output quality or fitness for purpose. User assumes all responsibility for use.

Template

Interpret the effect reported in {result_context} in terms of real-world magnitude, not just significance. (1) Identify the effect size metric and convert it to something interpretable: a standardized mean difference, a risk ratio or difference, a correlation, or better, the effect in the original natural units the audience understands. (2) Anchor the magnitude to a meaningful reference: an established minimally important difference for this outcome, the effect of a known comparator, or the natural variation in the outcome — a "large" statistical effect can be practically trivial and vice versa. (3) Compute an absolute, not just relative, expression: a 50% relative risk reduction on a rare event may be a tiny absolute benefit, and reporting only the relative figure misleads — provide number-needed-to-treat or absolute risk difference where applicable. (4) Report the confidence interval and interpret its full width, since the point estimate alone hides whether the effect could plausibly be negligible or huge. (5) Consider the effect's distribution: is it uniform, or driven by a responsive subgroup? (6) State a plain-language verdict on whether the effect matters for the decision at hand, separating the statistical question from the practical one.

Interpret the effect reported in {result_context} in terms of real-world magnitude, not just significance. (1) Identify the effect size metric and convert it to something interpretable: a standardized mean difference, a risk ratio or difference, a correlation, or better, the effect in the original natural units the audience understands. (2) Anchor the magnitude to a meaningful reference: an established minimally important difference for this outcome, the effect of a known comparator, or the natural variation in the outcome — a "large" statistical effect can be practically trivial and vice versa. (3) Compute an absolute, not just relative, expression: a 50% relative risk reduction on a rare event may be a tiny absolute benefit, and reporting only the relative figure misleads — provide number-needed-to-treat or absolute risk difference where applicable. (4) Report the confidence interval and interpret its full width, since the point estimate alone hides whether the effect could plausibly be negligible or huge. (5) Consider the effect's distribution: is it uniform, or driven by a responsive subgroup? (6) State a plain-language verdict on whether the effect matters for the decision at hand, separating the statistical question from the practical one.

Variables

Name	Type	Required	Trust level
result_context		yes

effect-sizeclinical-significancestatisticschain-of-thought

Ratings

0.0

Overall (0)

0.0

Accuracy

0.0

Consistency

0.0

Clarity

0.0

Efficiency

Benchmarks

Not yet self-validated against any benchmark. Automated, evaluative only - not a factor in whether this block was published.

Submitted by James P FounderMod via mcp · 2026-07-18