摘要：Paired samples Configuration Frequency Analysis (CFA) introduced by Stemmler and
Hammond (1997) is suggested as a valuable method for the multivariate detection
of change in small sample designs. We examined the assumption that impulsive
behavior in a Go/NoGo task, which is bivariately defined by increased speed and
decreased accuracy, is not only determined by personality traits like
impulsivity, but can also be manipulated by reinforcement conditions. By
applying paired samples CFA we could show that impulsive behavior could be
induced by positive reinforcement (reward), as well as suppressed by indirect
punishment (loss of reward). We could confirm this result parametrically with an
analysis of variance for repeated measures. A significant multivariate treatment
effect on reaction latency and passive avoidance errors was obtained. We argue
that paired sample CFA is an appropriate method for experimental and clinical
small sample designs, which can be applied to data of all dignity levels (e.g.
response latencies, rating scales, syndrome categories), does not imply
assumptions about sample distributions, following a case oriented approach,
which means that cases contributing to the significant effects can be directly
identified from the cross-tabulation. In conclusion, paired samples CFA can be
considered as a transparent, easy to interpret method for detecting changes in
small sample within-subjects designs.
关键词：Configuration Frequency Analysis (CFA), paired samples, multivariate analysis of
change, small samples
