ABSTRACT
In this chapter we describe a wide range of response processes, producing the following types of observed responses: • Continuous or metric • Dichotomous • Grouped • Censored • Ordinal • Unordered polytomous or nominal • Pairwise comparisons • Rankings or permutations • Counts • Durations or survival The aim of statistical modeling is to capture the main features of the empir-
ical process under investigation (see Section 8.2 for further discussion). Typically, a first simplifying step is to focus on a restricted set of response variables and to consider the data generating process of these variables given a set of explanatory variables. Univariate models have one response variable whereas multivariate models have several, possibly including intervening or intermediate variables serving as both response and explanatory variables. The response variables are sometimes called ‘dependent’, ‘endogenous’ or ‘outcome’ variables whereas the explanatory variables are called ‘independent’, ‘exogenous’ or ‘predictor’ variables. The explanatory variables of primary interest are sometimes called ‘exposures’ or ‘(risk) factors’ and the others ‘confounders’ or ‘covariates’. However, the term ‘covariate’ is often used as a generic term for explanatory variable.
