Suchergebnisse

Statistics is the one discipline in which people have concentrated almost exclusively on how to collect and analyze data. Statisticians have been doing this as a goal in itself without having to be encumbered by substantive issues arising from some other discipline. Most of their work can be broken into components that fall in one of three categories: data gathering, data reduction (descriptive statistics), and statistical inference.

Though many reasons have been suggested as to why the party division of the presidential & congressional vote in the US remains in 50% equilibrium, the historical evidence of the existence of equilibrium tendencies has never been tested. How improbable is it that the 2-party division of the vote would have remained within the region in which it has moved since the Civil War if it were governed by historical processes from which restoring forces were completely absent? Applying the mathematical theory of random walks, a test is made of the restoring-force hypothesis by calculating the probability that the party division would have stayed within its historic boundaries if it had moved as a symmetric random walk in one dimension. Except for the requirement of symmetry about the present division of the vote, the probability laws governing the movement of the party division in these models were fitted as closely to historical reality as possible. The results of the test indicate that restoring forces have in fact been present in US party competition The 0- hypothesis that the party division could have been governed by processes from which such forces are absent can be rejected at a high level of confidence. AA.

Der Autor geht davon aus, daß die in der Kontextanalyse verwendeten statistischen Gleichungen ebenso wie in der allgemeinen Analyse den in der Wirklichkeit vorfindbaren Prozeß reflektieren sollen, der in erster Linie die Daten auch generiert. Daraus folgt, daß vor der Modellspezifikation der eigenständige Prozeß verstanden werden muß. Die Kontextanalyse arbeitet deshalb mit einem Konzept von Variableneffekten sowohl auf dem individuellen als auch dem gruppenspezifischen Niveau, die jeweils zu unterscheidende Effekte auf der Ebene des Gegenstands, nicht auf der statistischen Ebene, evozieren. Der Beitrag führt in die Hauptfragen der Kontextanalyse ein und behandelt darüber hinaus drei spezifische Probleme: (1) Formen der Beeinflussung zweier Variablen; (2) Meßbarkeit des Beeinflussungseffekts und (3) Möglichkeiten des Rückschlusses auf Individualdaten von beobachteten gruppenspezifischen Aggregatdaten. (HN)