Issue 4
Vol. 42
2000
|
|
|
|
Optimization
of adaptive testing Because of lack of practical experiences with adaptive testing it is not known how reliability requirements will determine the number of items to be tested. In adaptive testing items are assembled according to number and quality in an optimal fashion for any participant in order to assure a pre set measurement error. It is vital to know, how many items will be tested in a population in regard to their frequency level and spread. In the following simulation study it is shown that for the two parameter logistic test model with fixing overall reliability at .80 there are on the average 10 items to be tested in a measurement oriented approach. However, on the average 35 items will be used with a reliability requirement of .95. That there are three and a half times more items to be tested is to be taken into account while planning for high levels of reliability. On the contrary, in a decision oriented approach the number of items can be much smaller as long as the decision is not altered by the adaptive decision making. Aparat from some extreme cases, there are six items to be tested on the average in regard to a preset cutoff and decision risk level. Key words: adaptive testing, error of measurement, diagnostic decision making Psychologische Beiträge, Volume 42, 4/2000 Prof. Dr. Lutz F. Hornke
|
