PURPOSE: The present work describes how vocabulary ability as assessed by 3 different forms of the Peabody Picture Vocabulary Test (PPVT; Dunn & Dunn, 1997) can be placed on a common latent metric through item response theory (IRT) modeling, by which valid comparisons of ability between samples or over time can then be made. METHOD: Responses from 2,625 cases in a longitudinal study of 697 persons for 459 unique PPVT items (175 items from Peabody Picture Vocabulary Test--Revised [PPVT-R] Form M [Dunn & Dunn, 1981], 201 items from Peabody Picture Vocabulary Test--3 [PPVT-3] Form A [Dunn & Dunn, 1997], and 83 items from PPVT-3 Form B [Dunn & Dunn, 1997]) were analyzed using a 2-parameter logistic IRT model. RESULTS: The test forms each covered approximately ± 3 SDs of vocabulary ability with high reliability. Some differences between item sets in item difficulty and discrimination were found between the PPVT-3 Forms A and B. CONCLUSIONS: Comparable estimates of vocabulary ability obtained from different test forms can be created through IRT modeling. The authors have also written a freely available SAS program that uses the obtained item parameters to provide IRT ability estimates given item responses to any of the 3 forms. This scoring resource will allow others with existing PPVT data to benefit from this work as well.
PURPOSE: The present work describes how vocabulary ability as assessed by 3 different forms of the Peabody Picture Vocabulary Test (PPVT; Dunn & Dunn, 1997) can be placed on a common latent metric through item response theory (IRT) modeling, by which valid comparisons of ability between samples or over time can then be made. METHOD: Responses from 2,625 cases in a longitudinal study of 697 persons for 459 unique PPVT items (175 items from Peabody Picture Vocabulary Test--Revised [PPVT-R] Form M [Dunn & Dunn, 1981], 201 items from Peabody Picture Vocabulary Test--3 [PPVT-3] Form A [Dunn & Dunn, 1997], and 83 items from PPVT-3 Form B [Dunn & Dunn, 1997]) were analyzed using a 2-parameter logistic IRT model. RESULTS: The test forms each covered approximately ± 3 SDs of vocabulary ability with high reliability. Some differences between item sets in item difficulty and discrimination were found between the PPVT-3 Forms A and B. CONCLUSIONS: Comparable estimates of vocabulary ability obtained from different test forms can be created through IRT modeling. The authors have also written a freely available SAS program that uses the obtained item parameters to provide IRT ability estimates given item responses to any of the 3 forms. This scoring resource will allow others with existing PPVT data to benefit from this work as well.
Authors: Jacob J Oleson; Joseph E Cavanaugh; J Bruce Tomblin; Elizabeth Walker; Camille Dunn Journal: Stat Methods Med Res Date: 2014-05-11 Impact factor: 3.021