TY - JOUR
T1 - A sampling and classification item selection approach with content balancing
AU - Chen, Pei-Hua
PY - 2015/1/1
Y1 - 2015/1/1
N2 - Existing automated test assembly methods typically employ constrained combinatorial optimization. Constructing forms sequentially based on an optimization approach usually results in unparallel forms and requires heuristic modifications. Methods based on a random search approach have the major advantage of producing parallel forms sequentially without further adjustment. This study incorporated a flexible content-balancing element into the statistical perspective item selection method of the cell-only method (Chen et al. in Educational and Psychological Measurement, 72(6), 933–953, 2012). The new method was compared with a sequential interitem distance weighted deviation model (IID WDM) (Swanson & Stocking in Applied Psychological Measurement, 17(2), 151–166, 1993), a simultaneous IID WDM, and a big-shadow-test mixed integer programming (BST MIP) method to construct multiple parallel forms based on matching a reference form item-by-item. The results showed that the cell-only method with content balancing and the sequential and simultaneous versions of IID WDM yielded results comparable to those obtained using the BST MIP method. The cell-only method with content balancing is computationally less intensive than the sequential and simultaneous versions of IID WDM.
AB - Existing automated test assembly methods typically employ constrained combinatorial optimization. Constructing forms sequentially based on an optimization approach usually results in unparallel forms and requires heuristic modifications. Methods based on a random search approach have the major advantage of producing parallel forms sequentially without further adjustment. This study incorporated a flexible content-balancing element into the statistical perspective item selection method of the cell-only method (Chen et al. in Educational and Psychological Measurement, 72(6), 933–953, 2012). The new method was compared with a sequential interitem distance weighted deviation model (IID WDM) (Swanson & Stocking in Applied Psychological Measurement, 17(2), 151–166, 1993), a simultaneous IID WDM, and a big-shadow-test mixed integer programming (BST MIP) method to construct multiple parallel forms based on matching a reference form item-by-item. The results showed that the cell-only method with content balancing and the sequential and simultaneous versions of IID WDM yielded results comparable to those obtained using the BST MIP method. The cell-only method with content balancing is computationally less intensive than the sequential and simultaneous versions of IID WDM.
KW - Automated test assembly
KW - Flexible content balancing
KW - Multiple forms
KW - Sampling and classification method
UR - http://www.scopus.com/inward/record.url?scp=84895115906&partnerID=8YFLogxK
U2 - 10.3758/s13428-014-0452-4
DO - 10.3758/s13428-014-0452-4
M3 - Article
C2 - 24610145
AN - SCOPUS:84895115906
SN - 1554-351X
VL - 47
SP - 98
EP - 106
JO - Behavior Research Methods
JF - Behavior Research Methods
IS - 1
ER -