Skip to main content

Advertisement

Table 1 Support for model structures. Model A was the best-supported model in italics. Changes in structures from model A are indicated. The differences in WAIC compared to model A were greater than 13 for all models, except model J. Given that model J was more complex, that model A was nested within model J, and that the WAIC difference was small (less than 2), the simpler model A was accepted for parsimony

From: Multiple-true-false questions reveal more thoroughly the complexity of student thinking than multiple-choice questions: a Bayesian item response model comparison

  Model structures WAIC ΔWAIC WAIC SE P WAIC
A Mastery, TTFF partial mastery, informed reasoning based on attractiveness, informed reasoning with double-T endorsement bias, individual student performance 18,520.7 0 200.0 342.1
B − remove question-level mastery 18,927.0 − 406.3 199.1 288.7
C − remove TTFF partial mastery 18,566.0 − 45.3 200.2 328.9
D − remove TTFF partial mastery + replace with TTFF-TFTF partial mastery 18,534.3 − 13.6 199.7 341.8
E − remove TTFF partial mastery + replace with TTFF-TFTF-TFFT partial mastery 18,536.5 − 15.8 199.5 333.5
F − remove informed reasoning based on attractiveness + replace with random guessing 19,582.4 − 1061.7 203.4 268.4
G − remove double-T bias 18,656.1 − 135.4 201.4 330.3
H − remove double-T bias + replace with multi-T bias 18,533.4 − 12.7 200.3 344.6
I − remove question-level double-T bias + replace with global double-T bias for each student 18,539.6 − 18.9 200.0 325.1
J + add random guessing for students not in mastery, partial mastery, or informed reasoning 18,519.9 + 0.8 199.9 356.1
K − remove individual student performance 19,448.5 − 927.8 196.7 180.2