Mediation and Causality at the Individual Level

摘要

Within a person-oriented research paradigm the focus is on individuals characterized by patterns of information that are regarded as indivisible wholes. It is then not sufficient to carry out standard variable-oriented mediation analysis. The procedure suggested by von Eye, Mun, and Mair (2009) for pattern-oriented mediation analysis is much better aligned to this person-oriented framework. An important new feature in their approach is that it can detect mediator configurations that prohibit predictor and outcome connections at a pattern level. Two extensions of their procedure are suggested, namely (1) the use of cluster analysis to arrive at the categories and (2) the use of other models for estimating the expected frequencies. It is pointed out that in their context a functional relations perspective might be more relevant than the standard causality perspective.