Optimal design for online generated intelligence tests

点击翻译按钮获取中文摘要

In the first and second phase of this project rule-based items measuring reasoning and mental speed have been developed. These items which are generated on the fly have been calibrated using linear logistic test models and Rasch Poisson counts models. To efficiently estimate the parameters optimal designs for both models including fixed as well as random effects have been derived. Furthermore, an extensible test platform based on R is provided as open source software.The developments of the first and second phase are to be continued in the third phase. First, rule-based items to measure memory capacity will be created and integrated in the test platform. Thus, all components of the Berlin intelligence structure model besides divergent thinking can be measured using rule-based items on the fly. Second, the construct validity of rule-based items measuring reasoning will be analysed when these items are applied at several times. Further dependent variables to be studied are the within-family variance of the items parameters, the invariance of these parameters and the growth curve of the performance.