Abstract (eng)
In a 2AFC-Experiment, participants are usually asked to respond as fast and as accurate as possible. This instruction results in a conflict of strategies: If a participant responds accurately, the velocity of the response decreases. If a participant responds fast, wrong reactions are made more frequently. Wrong reactions are often excluded from the analysis f.e. of average reaction times (= reduced-sample estimator). From a statistical point of view, a wrong reaction results in a censoring problem, which could be resolved with the Kaplan-Meier method or an adapted version of the Aalen-Johansen estimator. Wrong reactions are sometimes caused by guessing trials. If guessing trials are subtracted from the data, theoretically, the “true” performance should be estimated even more accurately. The Kaplan-Meier method, the Aalen-Johansen estimator and a combination of both with the Kill-the-Twin procedure should lead to more accurate, hence more reliable estimations of median reaction times. A tendency towards more reliable estimations for wrong-reaction-corrections in combination with guessing corrections was found, followed by more reliable estimations for wrong-reaction-corrections compared to the reduced-sample estimator.