Table 3. OLS Regression Output.
|
Basic model:
Border predictor only |
Parsimonious model:
predictors based on IC |
Hypothesis-driven model:
all available predictors |
||||
|---|---|---|---|---|---|---|
| Diagnostics | ||||||
| Mean VIF | - | 1.7 | 3.0 | |||
| BIC | 204 | 196 | 199 | |||
| R 2 | 0.25 | 0.56 | 0.77 | |||
| Results | Coeff. | 95% CI | Coeff. | 95% CI | Coeff. | 95% CI |
| Border county | 15.7 | 4.3-27.2 | 19.5 | 10.0 to 29.0 | 16.7 | 6.4 to 26.9 |
| CV deaths | -120.9 | -243.5 to 1.7 | -123.9 | -267.4 to 19.7 | ||
| Pop density | 5.5 | 0.7 to 10.4 | 11.2 | 4.0 to 18.4 | ||
| Pop ageing | 229.5 | -186.5 to 645.5 | ||||
| Female | -545.0 | -1963.7 to 873.7 | ||||
| NH beds | 0.2 | -0.3 to 0.7 | ||||
| Respiratory deaths | -70.1 | -344.0 to 203.8 | ||||
| Live alone | -193.0 | -519.0 to 133.1 | ||||
| Social class | -183.8 | -327.6 to -40.1 | ||||
Legend:VIF: variance inflation factor, a measure of collinearity where lower is better. BIC: Bayesian information criterion, a measure of information loss where lower is better. R 2: a measure of goodness of fit, where higher is better. Coeff.: coefficient. The estimated change in outcome with a one unit increase in the predictor, holding all other factors in the model constant. CI: confidence interval. Border county=1 if county borders Northern Ireland. All other predictors: see Table 1. Statistically significant (p<0.05) results highlighted bold.