Hypothesis
Because Registered Reports are preregistered and often share data and code, their main results should be independently reproducible from the shared materials.
What failed
Of 62 Registered Reports published 2014–2018, 41 had data and 37 had scripts available, 36 had both, 31 scripts would run, and the main results were reproduced for only 21 — 58% of those that shared both.
Broken assumption
Sharing data and analysis code is enough to make published analyses reproducible.
Why
The authors attribute the shortfall to data and code that are not shared, scripts that will not run, and data annotated too poorly to tell what variables and values mean, noting most researchers are not trained in computational reproducibility.
From the published paper: Obels et al. Advances in Methods and Practices in Psychological Science, 2020. ↗CC BY