Reading Comprehension Tests for Children: Test Equating and Specific Age-Interval Reports

Abstract

Equating is used to directly compare alternate forms of tests. We describe the equating of two alternative forms of a reading comprehension test for Brazilian children (2nd to 5th grade), Form A (n = 427) and Form B (n = 321). We employed non-equivalent random groups design with internal anchor items. Local independence was attested via standardized residual Pearson's bivariate correlation. First, from 176 items, we selected 42 in each form (33 unique and 9 in common) using 2PL model, a one-dimensional item response theory (IRT) model. Using the equateIRT package for R, the anchor items were used to link both forms. Linking coefficients were estimated under two different methods (Haebara and Stocking–Lord), resulting in scores equating by two methods: observed score equating (OSE) and true score equating (TSE). We provided reference-specific age-intervals for the sample. The final version was informative for a wide range of theta abilities. We concluded that the forms could be used interchangeably.

Impact of OrCam Learn on reading comprehension

“…To compare reading comprehension outcomes, standardised tests were used. For the purposes of this study, as per Lúcio et al (2021) use of alternative/parallel forms within the same test battery is the preferred strategy to compare reading students' attainment, without needing to consider the effect of previous assessment on outcomes. When assessing reading comprehension, particularly twice in a short period of time, it is important to choose test protocols that do not risk compromising students' well-being.…”