Skip to main content
Dog park counter's dawn undercount does not reappear at five more sites

Dog park counter's dawn undercount does not reappear at five more sites

A new Slop University paper repeats an earlier dog park gate counter study at five further sites and does not recover the dawn undercount that five councils now correct for in their off-lead-hours decision rules.

A pre-registered replication across five metropolitan councils puts the dawn to midday agreement gap at 2.1 percentage points, against 33 in the study the councils built on

Slop University has gone back to one of its own published estimates and, on the evidence of five further sites, cannot find it again.

Five Gates, No Dawn Gap, released today through the Office of Research Outputs, is a pre-registered replication of an earlier School of Continuous Improvement finding that a suburban dog park’s automatic gate counter agreed with manual headcounts far less often at dawn than at midday. Five metropolitan councils have since adopted a shared dawn correction factor drawn from that result. Across the five new sites, the dawn-to-midday gap comes in at 2.1 percentage points, against 33 in the original, and only one site shows any attenuation at all.

The result we could not reproduce is still doing work in five decision rules, and that seemed to us the part worth putting on the record.

— Dr Petra Umbile, Postdoctoral Fellow, School of Emergent Priorities

The team held the original observation protocol fixed — paired observers, six two-hour blocks, twenty-four weeks at each enclosure — and lodged its analysis plan and equivalence bounds before the first count. Of 720 planned sessions, 694 were completed.

Professor Verity Marris, Director of the Trajectory Analytics Group, said the School had learned something about its own habits in the process: “We are a good deal quicker at putting an estimate into service than at going back to ask it a second question, and this has been a useful reminder of the gap.”

The University regards the re-examination of its own published estimates as part of what a research record is for. The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.n23z4k.