TY - JOUR
T1 - An item response theory–based, computerized adaptive testing version of the macarthur–bates communicative development inventory
T2 - Words & sentences (CDI:WS)
AU - Makransky, Guido
AU - Dale, Philip S.
AU - Havmose, Philip
AU - Bleses, Dorthe
PY - 2016/4/1
Y1 - 2016/4/1
N2 - Purpose: This study investigated the feasibility and potential validity of an item response theory (IRT)–based computerized adaptive testing (CAT) version of the MacArthur–Bates Communicative Development Inventory: Words & Sentences (CDI:WS; Fenson et al., 2007) vocabulary checklist, with the objective of reducing length while maintaining measurement precision. Method: Parent-reported vocabulary for the American CDI: WS norming sample consisting of 1,461 children between the ages of 16 and 30 months was used to investigate the fit of the items to the 2-parameter logistic IRT model and to simulate CDI-CAT versions with 400, 200, 100, 50, 25, 10, and 5 items. Results: All but 14 items fit the 2-parameter logistic IRT model, and real data simulations of CDI-CATs with at least 50 items recovered full CDI scores with correlations over.95. Furthermore, the CDI-CATs with at least 50 items had similar correlations with age and socioeconomic status as the full CDI:WS. Conclusion: These results provide strong evidence that a CAT version of the CDI:WS has the potential to reduce length while maintaining the accuracy and precision of the full instrument.
AB - Purpose: This study investigated the feasibility and potential validity of an item response theory (IRT)–based computerized adaptive testing (CAT) version of the MacArthur–Bates Communicative Development Inventory: Words & Sentences (CDI:WS; Fenson et al., 2007) vocabulary checklist, with the objective of reducing length while maintaining measurement precision. Method: Parent-reported vocabulary for the American CDI: WS norming sample consisting of 1,461 children between the ages of 16 and 30 months was used to investigate the fit of the items to the 2-parameter logistic IRT model and to simulate CDI-CAT versions with 400, 200, 100, 50, 25, 10, and 5 items. Results: All but 14 items fit the 2-parameter logistic IRT model, and real data simulations of CDI-CATs with at least 50 items recovered full CDI scores with correlations over.95. Furthermore, the CDI-CATs with at least 50 items had similar correlations with age and socioeconomic status as the full CDI:WS. Conclusion: These results provide strong evidence that a CAT version of the CDI:WS has the potential to reduce length while maintaining the accuracy and precision of the full instrument.
U2 - 10.1044/2015_JSLHR-L-15-0202
DO - 10.1044/2015_JSLHR-L-15-0202
M3 - Journal article
AN - SCOPUS:84964758651
SN - 1092-4388
VL - 59
SP - 281
EP - 289
JO - Journal of Speech, Language, and Hearing Research
JF - Journal of Speech, Language, and Hearing Research
IS - 2
ER -