ABSTRACT

As seen in Chapters 5 and 6, statistical modelling consists in general of relating an outcome (a dependent variable) to a series of predictors (sometimes also called explanatory variables). This dichotomy is so classic that it appears natural and we sometimes forget that in practice there can be two or three relevant outcomes that could be advantageously studied all together, or that a given predictor can be considered the outcome of another predictor. For instance, the “income” of a young adult can be explained in a linear regression model by the two explanatory variables “educational level” and “parental income.” But, at the same time, “educational level” can be explained by “parental income” (and not the reverse, due to a question of temporality). Unfortunately, it is not possible at the moment to implement? in a linear regression model an asymmetrical relationship of this sort between two predictors.