Answers are grounded in this paper and cite their sources.
Temporal variations in the incidence of rare diseases are a subject of great interest to epidemiologists, who look for recognizable patterns and associations with putative causal factors in order to gain aetiological clues. These data are typically characterised by overdispersion and correlated errors, nevertheless the standard method of analysis in this field has so far been the Poisson regression. The aim of this preliminary study was to suggest the novel employment of a Bayesian nonparametric model (based on a GLMM), a validated statistical approach, to the context of the estimation of childhood cancer incidence trends. The model includes fixed effect terms (age and year of diagnosis) and a random effect vector, intended to represent smooth variation over time, specified in the forward direction as a second order Gaussian autoregressive component. The best use for the model proposed is as "early warning" to epidemiologists for the identification of changes in trends. This and similarly flexible models are in fact characterised by high sensitivity to changes of behaviour; the price we must pay for such high a sensitivity being low specificity.
No takes yet. Share an insight, caveat, or question.
Zuccolo et al. (2005) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: