Supports· moderateMPI-E006
Grote 2017: preplanned analyses null; a post-hoc variant significant
- what the source states
- Twenty participants, 720,000 bits, three analyses defined and tested before looking at the data — including a conceptual replication of the correlation-matrix method. Results: p = 0.438, p = 0.703, and p = 0.0949 for the three preplanned analyses (combined p = 0.315). "While none of the pre-defined analysis results is found significant, a post-hoc variant of analysis 3 that includes the control data is significant with p = 0.012."
- what we infer
- One study, both lessons: the preplanned nulls extend the failed replication record (MPI-C002), and the significant post-hoc variant is a live demonstration of how analytic flexibility can conjure a positive from the same data (MPI-C008). The synthesis document that seeded this case cited the post-hoc figure as a replication success — an error this import corrects.
- limitations
- Small sample (20 participants) limits what the preplanned nulls can exclude.