Table 2.
Mouth cancer in Mumbai 1995–2009, men aged 25–74. Analysis of deviance for nested APC models.
| Model No. | Model description | Goodness-of-fit |
Model Comparison | Effect Tested | Difference between models |
||||
|---|---|---|---|---|---|---|---|---|---|
| dfa | Residual Deviance | p | dfa | Deviance | p | ||||
| 0 | Age | 9 | 149.8 | <0.0001 | |||||
| 1 | Age + drift | 10 | 122.3 | <0.0001 | 1 vs. 0 | Drift | 1 | 27.5 | <0.0001 |
| 2 | Age + period | 11 | 118.8 | <0.0001 | 2 vs.1 | Non-linear period | 1 | 3.5 | 0.0080 |
| 3 | Age + cohort | 20 | 107.7 | <0.0001 | 3 vs.1 | Non-linear cohort | 10 | 14.6 | 0.0012 |
| 4 | Age+ period + cohortb | 21 | 103.2 | <0.0001 | 4 vs. 3 | Non-linear periodc | 1 | 4.5 | 0.002 |
| 4 vs.2 | Non-linear cohortc | 10 | 15.6 | 0.0005 | |||||
Degrees of freedom.
Best-fitting APC model on the grounds of significant non-linear period and cohort effects.
Period (cohort) adjusted for non-linear cohort (period).