Table 2.

Summary of the different distributed lag non-linear models (DLNMs) characterizing the relationship between clinical incidence and ante natal clinic (ANC) parasite prevalence

Acronym	Endemicity effect	Lagged effect	Number of parameters	AIC	RMSE (rolling cross-validation)
LE	Linear	No lagged effects	6	3859.2	0.0667
LELL	Linear	Linear	7	3116.6	0.0563
LENL	Linear	Non-linear	13	3116.0	0.0564
NE	Hill function	No lagged effects	8	3499.8	0.1126
NELL	Hill function	Linear	9	2982.0	0.05434
NENL	Hill function	Non-linear	15	2978.9	0.05431

The second and third columns indicate the shape of the basis function used to characterize how the relationship is influenced by endemicity and the lagged effect. Models are compared using Akaike information criterion (AIC, lowest value in italic indicating most parsimonious model) and root mean squared error (RMSE, lowest value in italic indicating most predictive model)