Fig. 4

Meningioma incidence: model selection and estimable functions. A Composite goodness-of-fit tests [11] for proportionality (PH-L, PH-T, PH-X, PH-A) and EF parallelism (Par-LAC, parallel Longitudinal Age Curves; Par-CAC, parallel Cross-Sectional Age Curves; Par-FTT, parallel Fitted Temporal Tends; Par-FCP, parallel Fitted Cohort Patterns). Abscissa values show \(-\log_{10}\) (P – Values). B Bias-corrected AIC values, \({AIC}_{c}\), relative to the model with lowest \({AIC}_{c}\). C – H Females, solid blue curves, Males, dashed red curves. Longitudinal Age Curves (C), Fitted Cohort Patterns (D), Cross-Sectional Age Curves (E), Fitted Temporal Trends (F), Local Drifts (G), and Gradient Shifts (H). All curves are estimated under the PH-T model. C, D, E, and F EF rates per 100,000 on the natural log scale. G and H EF annual percentage changes