| Abstract | Invited by the National Educational Panel Study (NEPS), during the Winter 2011/12, the author gave a series of lectures about ‘Statistical methods in sociological research of education’. This text comprises an elaboration of three of these lectures discussing the understanding of statistical models and their relationships with substantive research questions. A basic idea is to distinguish between two goals of statistical methods: Description of statistical facts concerning frequency distributions defined for a sample or population, and finding rules for relationships between variables which can be used for predictions and explanations. Correspondingly, there are two kinds of statistical generalization: Descriptive generalizations (from sample to population), and modal generalizations (from data to rules). The paper is then mainly concerned with functional models understood as tools for modal generalizations. Such models can be used both for predictions and for explanations. In social research, they are mainly used for explanations. Important questions then concern relationships between explanatory variables which can be of different kinds. In particular, one has to distinguish between interactions and functional relationships. Further questions concern causal interpretations. It is argued that causal relationships cannot be defined in terms of functional models (variables and their functional relations), but require an interpretation in terms of the substantive application. How to understand such interpretations also depends on whether the model is intended to serve explanations or to assess effects of treatments. This distinction is particularly important for social research dealing with processes involving activities of human agents. The paper finally discusses, and contrasts, two quite different understandings of notions of ‘potential outcomes’. (DIPF/Orig.) |
|---|
Verfügbarkeit