Adaptive Testing of the Number Series Test Using Standard Approaches and a New Decision Tree Analysis Approach
JOHN J. McARDLE · 2013
In this chapter the author attempts to demonstrate the utility of a Decision Tree Analysis (DTA) approach to Computerized Adaptive Testing (CAT). The basic psychometric premise used here is that if the overall (or total) score for any individual comes from a full set of items (I ), the behavior on a smaller number of items (i 1,200) were effectively administered all 47 items. Standard CAT strategies are compared to the DTA approach and we conclude that the DTA is much more accurate with a scale reliability of ρ2 = 0.85 with only 4-7 items administered. Other costs versus benefits type experiments are suggested.