Table 3.

ANOVA from nested GAMs with non-linear time trend

Modela	p value	Adjusted deviance explained (D2 adj)b
1. Year		0.09
2. Year + age at onset	<0.001	0.14
3. Year + age at onset + sex	0.28	0.14
4. Year + age at onset + sex + country	<0.001	0.81
5. Year + age at onset + sex + country + year × age at onset	0.001	0.81

^aThe models are evaluated with the response variable being the incidence of type 1 diabetes and the log of population at risk as the offset. For the time trend, we use a smooth nonparametric term; linear coefficients are assumed for age, sex and interaction terms

^bD² adj is the equivalent of R² adjusted for generalised linear models, taking into account the number of observations and the number of predictors, and thus permitting direct comparison among models