Dunning: DECEPTIVE PRACTICE — Crime reports

• Exploratory description, not a forecast, causal explanation or proof of a change.
• Approximate 95% factor intervals assume independent, constant-variance log residuals and fixed boundaries. They do not account for boundary search, serial dependence or source revisions.
• Confirmed zeros are treated as 1 only for fitting and its diagnostics. Raw counts stay zero; low-count results depend on this convention.
• Independent phase lines may jump at boundaries. Every observation belongs to exactly one phase.
• Reporting, classifications and seasonality can change. Counts are not population-adjusted; a level shift without a slope change is not the target of this search.
• Uses dated last-good cached data, not newly verified source counts.
• Boundary ranges show near-optimal models within 2 BIC units, not confidence intervals. They can be narrow even when the model is uncertain.
• At least one factor interval includes 1; its direction is uncertain under this approximation.
• Many dependent filter conditions were searched. No multiple-testing-adjusted significance or causal attribution is claimed.
• Frequency compares fixed equal-length windows of raw counts, not population risk or a celeration-factor ratio. Reporting, classifications and exposure may have changed.

Intervals assume fixed boundaries and independent log residuals. They are not adjusted for the phase or filter search. Near-optimal boundary ranges are not confidence intervals.

Near-optimal boundary range: October 2020–October 2020.

Near-optimal boundary range: January 2022–March 2022.

Newest first. Unknown coverage is not a zero. These are the original saved observations, not fitted or rounded predictions.

Publisher dataset

Crimes — 2001 to Present · City of Chicago

Source retrieved

2026-11-05T21:08:01.557Z

Publisher update

2026-11-05T12:12:19.000Z

Covered months

October 2016–September 2026

Source note

Source ijzp-q8t2. Joint day/category/hour/applicable-flag/ward/status counts; no addresses, coordinates or incident identifiers. Each year's cells reconcile with an independent source total. Reads are not an atomic publisher snapshot. Published crime reports, not population-adjusted rates or final incidence. Prepared daily community-area snapshot; scheduled rebuild at 06:00 America/Chicago. Full supported history is rechecked for corrections. Retrieval timestamps are preserved; this is not a real-time feed. Crime analysis starts January 1, 2002; 2001 is excluded from counts and fitting inputs. Local indexed neighborhood snapshot. Totals are for the loaded area before other filters, not citywide totals. Citywide and unassigned-area views use a separate network-backed scope.

Snapshot

daily-v1:ef0c5e9b-09d5-4a2d-849e-1be6745dbfbb:neighborhood-v2:d493ad0a803f1d54ea6c30cd03756edc

Discovery method

multi-feed-turn-and-frequency-renewal-v2

Analyze method

exploratory-segmented-log10-ols-zero-as-one-v3

Search plan

a3ffccedeb8d8ce18a30df3636839f775af1e5188e55a1f685695133135c6e35

Evidence run

4461f10f5ac01c63aa4d6ac56d8c246a12c952370ca342e3898af0d8b6e975ab