Experimental Reproducibility Diagnosis Chain

validated scientific chain v1.0.0 cc-by-sa

Block ID: a8da4b73-4777-4c39-9d25-8e34a96fbab3

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

Template

Diagnose why the experiment in {replication_context} failed to reproduce, working from most to least likely cause. (1) Confirm the failure is real: is the replication result statistically distinguishable from the original, or merely a non-significant point estimate in the same direction? Underpowered replications "fail" by design — check the replication's own power first. (2) Compare protocols line by line: reagent lots and suppliers, equipment models and calibration, environmental conditions, operator technique, timing, and software versions — the difference is usually a hidden variable the original never reported because it wasn't known to matter. (3) Scrutinize the original for fragility: was it a single significant result among many tests, a small sample, or a p-value hovering near threshold? Such findings often reflect the original's luck, not the replication's failure. (4) Consider genuine effect heterogeneity: does the effect depend on a population, context, or moderator that differs between the two runs? A context-dependent effect is not a false one. (5) Check both directions for error — the replication can be flawed too. (6) Conclude with the most probable explanation and the decisive test: the one controlled variable to hold fixed that would confirm it.

Diagnose why the experiment in {replication_context} failed to reproduce, working from most to least likely cause. (1) Confirm the failure is real: is the replication result statistically distinguishable from the original, or merely a non-significant point estimate in the same direction? Underpowered replications "fail" by design — check the replication's own power first. (2) Compare protocols line by line: reagent lots and suppliers, equipment models and calibration, environmental conditions, operator technique, timing, and software versions — the difference is usually a hidden variable the original never reported because it wasn't known to matter. (3) Scrutinize the original for fragility: was it a single significant result among many tests, a small sample, or a p-value hovering near threshold? Such findings often reflect the original's luck, not the replication's failure. (4) Consider genuine effect heterogeneity: does the effect depend on a population, context, or moderator that differs between the two runs? A context-dependent effect is not a false one. (5) Check both directions for error — the replication can be flawed too. (6) Conclude with the most probable explanation and the decisive test: the one controlled variable to hold fixed that would confirm it.

Variables

Name	Type	Required	Trust level
replication_context		yes

reproducibilityreplication-failuremethodschain-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