Theory-Based Automatic Item Generation of Elementary and Middle School Area and Surface Area Problems

• In Event: Marriott Roundtable Session 48
  In Roundtable Session: Cognition and Assessment SIG Roundtable

Abstract

Strong-theory automatic item generation (AIG) of elementary- and middle-school area and surface area tasks was conducted. Feature specifications were developed for the content, and mathematical combinations of feature levels were created. Some sets of combinations were trimmed via D-optimal sampling. Resulting combinations were spiraled into formsets – 2 for grade 3, 8 for grade 6, and 2 for grade 7. 50 forms were generated for each formset. Given time constraints, only grade 3 content was tested. Approximately 240 rising fourth graders took the assessment online. Multilevel Rasch modeling was applied with the features used in AIG and other important variables as predictors. Between-item family variance accounted for approximately 38% of the total variance. Implications of results are discussed.

Authors

• Mary Ann Simpson

• Lisa Bickel, MetaMetrics

• Ian F Hembry, MetaMetrics

• Ruth Price

• Cassandra Totten

• Eleanor E. Sanford-Moore, MetaMetrics

• Heather Hughes Koons, MetaMetrics

• Pan Qianqian