Table 2.
Estimates of the effect of the QOF on health behaviours using local linear regression.
| BMI |
Alcohol |
No. cigarettes |
||||
|---|---|---|---|---|---|---|
| Coeff. | N. | Coeff. | N. | Coeff. | N. | |
| Model I: | ||||||
| With the optimal bandwidth | −0.22 (0.19) | 11,270 | −0.01 (0.02) | 9101 | −0.70** (0.28) | 19,663 |
| Model II: | ||||||
| With 1.5 times the optimal bandwidth | −0.18 (0.16) | 16,441 | 0.003 (0.02) | 14,119 | −0.67*** (0.23) | 27,269 |
| Model III: | ||||||
| With twice the optimal bandwidth | −0.09 (0.13) | 21,397 | 0.01 (0.02) | 17,883 | −0.62*** (0.21) | 32,102 |
| Model IV: | ||||||
| With 0.3 times the optimal bandwidth | −0.41 (0.35) | 3482 | 0.01 (0.04) | 2415 | −0.84* (0.49) | 5500 |
| Model V: | ||||||
| With 0.4 times the optimal bandwidth | −0.26 (0.31) | 4660 | 0.001 (0.04) | 3203 | −0.83* (0.43) | 7628 |
Note: optimal bandwidths are as follows: for BMI hopt = 2.5 years; for alcohol hopt = 1.9 years; for cigarettes hopt = 3.4 years. Std. errors in (). Weighted sample. All equations control for age and gender.***p < 0.01; **p < 0.05; *p < 0.1.