Negative binomial regression comparing cannabis stores per capita within 1000 m of the geographic centre of neighbourhoods in the lowest income quintile (Q1) and the highest income quintile (Q5)
| Retail model; jurisdiction | Incident rate ratio (95% CI), Q1:Q5 (reference) | |
|---|---|---|
| Model A* | Model B* | |
| Private/hybrid | ||
| Yukon Territory (hybrid) | –† | –† |
| British Columbia (hybrid) | 2.73 (1.83–4.06) | 2.02 (1.34–3.05) |
| Alberta | 2.89 (2.16–3.88) | 2.42 (1.79–3.27) |
| Saskatchewan | 2.27 (1.26–4.07) | 1.75 (0.97–3.17) |
| Manitoba | 4.89 (2.71–8.84) | 3.04 (1.66–5.58) |
| Ontario | 1.44 (1.02–2.05) | 1.02 (0.71–1.48) |
| Newfoundland and Labrador | 2.47 (1.36–4.51) | 1.32 (0.74–2.34) |
| Total | 2.64 (2.22–3.14) | 2.33 (1.96–2.79)‡ |
| Government | ||
| Northwest Territories | 0.63 (0.10–3.75) | 0.66 (0.11–3.97) |
| Quebec | 4.24 (2.57–7.00) | 2.84 (1.65–4.91) |
| New Brunswick | 2.41 (1.22–4.78) | 1.72 (0.84–3.52) |
| Nova Scotia | 4.63 (2.14–10.04) | 3.74 (1.73–8.16) |
| Prince Edward Island | 3.89 (0.82–18.33) | 1.66 (0.27–9.99 |
| Total | 3.50 (2.50–4.91) | 3.05 (2.16–4.31)‡ |
| Grand total | 2.71 (2.32–3.16) | 2.33 (1.98–2.72) |
Note: CI = confidence interval.
↵* Both models are offset by the total population of a dissemination area. Model B adjusts for the population density of each dissemination area in people per square kilometre.
↵† Unstable regression model, as income has near-perfect prediction for the dependent variable.
↵‡ Interaction between private/hybrid retail system v. government retail system and income quintiles was not significant (p = 0.2).