{"status":"success","data":[{"release":"2026.1","tag":"civilytics-crdc-arrests-2026.1","is_latest":true,"years":["15-16","17-18","21-22","23-24"],"default_model":"unified_m2_mod","notes":"An update to 2025.1 that adds 2023-24. It is preliminary: 2025.1 was vetted for the white paper, and 2026.1 has not yet been vetted the same way. Cite the white paper (Knowles & Miller 2025) and name the release you used. Numbers for 2015-16, 2017-18 and 2021-22 also moved, for two reasons. First, 2025.1 counted a school that did not report arrests (CRDC reserve codes such as -5, action plan, and -6, force certification) as a school with no arrests; such a school is now missing, and leaves both the arrest count and the enrollment it is divided by. Second, New York City, absent from 2025.1, now appears in 2021-22 and 2023-24 as its 33 geographic districts plus a citywide row (leaid 3620580, is_aggregate true). m1 and m2 now fit 2023-24 alone, so the default model, unified_m2, has no earlier years: request m3, m4 or m5 for those, or pin release=2025.1. m3 to m5 pool all four waves. Five district intercepts in unified_m3 have R-hat between 1.010 and 1.012. The other nine fits peak at R-hat 1.010 to 1.022, with minimum bulk ESS from 183 to 596; the default, unified_m2, has 21 parameters above 1.01. The data dictionary has the convergence diagnostics and the outlier notes."},{"release":"2025.1","tag":"civilytics-crdc-arrests-2025.1","is_latest":false,"years":["15-16","17-18","21-22"],"default_model":"unified_m2_mod","notes":"First release: estimates for 2015-16, 2017-18 and 2021-22."}],"error":null,"meta":{"total":2,"latest":"2026.1","data_release":"2026.1","version":"v1"}}