Controlling individuals' time spent on task in speeded performance measures. Experimental time limits, posterior time limits, and response time modeling.

Autor/inn/en	Goldhammer, Frank; Kröhne, Ulf
Titel	Controlling individuals' time spent on task in speeded performance measures. Experimental time limits, posterior time limits, and response time modeling.
Quelle	In: Applied psychological measurement, 38 (2014) 4, S. 255-267 PDF als Volltext kostenfreie Datei (1); PDF als Volltext (2) Link als defekt meldenVerfügbarkeit
Sprache	englisch
Dokumenttyp	online; gedruckt; Zeitschriftenaufsatz
ISSN	0146-6216; 1552-3497
DOI	10.1177/0146621613517164
URN	urn:nbn:de:0111-pedocs-127839
Schlagwörter	Empirische Untersuchung; Reliabilität; Validität; Leistungsdruck; Adaptives Testen; Item-Response-Theory; Testaufgabe; Schuljahr 12; Schüler; Schülerleistung; Technologiebasiertes Testen; Zeit; Leistungsmessung; Modell; Deutschland
Abstract	The speed-ability trade-off becomes a measurement problem if there is between-subject variation in the speed-ability compromise, as this may affect the comparability of ability estimates. To control individual speed differences, the response-signal (RS) paradigm was applied requiring an immediate response as soon as an acoustic signal is presented. A figural discrimination task and a word recognition task were completed both in an untimed condition allowing individual differences in time spent on task and in several timed conditions where the time available for item completion was limited using the RS paradigm. Thus, speed was manipulated by varying the available time between stimulus-onset and RS. A total of N = 205 high school students participated in the study. Results showed that across timed conditions with decreasing time on task, the ability level and ability variance decreased substantially. Ability correlations between timed conditions were high, whereas correlations between untimed and timed conditions were low. This finding suggested that ability differences being inconsistent to those found in the timed condition are due to individual differences in time on task in the untimed condition. To eliminate these differences, two ways were considered. First, untimed responses were recoded using two-tailed posterior time limits. As expected, correlations between timed and untimed conditions were increased. Second, the log-transformed item response times were included in the item response model, which led to even higher correlations between timed and untimed conditions. Validity and generalizability of the proposed testing procedure are discussed.
Erfasst von	DIPF \| Leibniz-Institut für Bildungsforschung und Bildungsinformation, Frankfurt am Main
Update	2017/3