Table 3. Characteristics of Persistent Cluster Area A, Cluster-Detection Analysis of Late-Stage Diagnosis of Colorectal Cancer, by Method, Scale, and Aggregation, Florida, 2006–2010a .
| Characteristic | Black | Cuban | White Hispanic |
|---|---|---|---|
| Significant at P <.10 | |||
| No. of scales | 11 of 12 | 0 of 12 | 11 of 12 |
| Unit of aggregation (block group or census tract) | Both | Neither | Both |
| Method used (Bernoulli or Poisson) | Both | Bernoulli | Both |
| Area selected based on | |||
| Scale, % of population | 40 | NA | 50 |
| Aggregation (unit of analysis) | Census tract | NA | Census tract |
| Method used | Poisson | NA | Poisson |
| Relative risk | 1.53 | NA | 1.36 |
| P value | .03 | NA | <.001 |
| County | Miami-Dade | NA | Miami-Dade and Broward |
| No. of late-stage cases | 197 | NA | 1,652 |
| Demographics | |||
| Population total in 2010 | 17,036 | NA | 72,967 |
| Hispanic, % | 14 | NA | 17 |
| Non-white, % | 50 | NA | 54% |
| Below poverty, % | 40 | NA | 31% |
a
Selection of area of geographic interest was based on P value, magnitude of risk, overlap, and evaluation of other persistent, significant clusters at that scale. Tract-level aggregation was selected to match with available area-based, sociodemographic information.