ABSTRACT
This chapter describes the likelihood for the standard self-controlled case series (SCCS) model, in which age and exposure effects are piecewise constant, that is, constant on intervals, because it is the most commonly used. The parameter estimates are the values that maximise the likelihood; this estimation method is known as maximum likelihood. The key difference is in italics: the SCCS likelihood conditions on the number of events observed for each individual. This implies that, in the SCCS method, information on the degree of association between exposure and event is obtained from the relative timing of events and exposures within individuals, rather than from marginal event counts, which are fixed by the conditioning. Rates can then be contrasted, adjusting for age effects and covariates using suitable statistical models, yielding estimates of rate or hazard ratios with direct interpretations as ratios of absolute incidences.
