Validation of the Preschool Child Observation Record: Does It Pass the Test for Use in Head Start?

Abstract

Research Findings: This study provides the first independent investigation of the second most widely used multidimensional assessment in Head Start—the Preschool Child Observation Record, Second Edition (COR-2). We conducted a comprehensive investigation into the validity of the COR-2 using data from all children in an urban school district's Head Start program (N = 4,071). Confirmatory factor analysis revealed a misfit between the 6 developer-defined categories and the data. Although exploratory analyses revealed a possible 4-factor solution, subsequent analyses indicated problems with this structure as well. Item response theory methods were used to determine whether there was support for the 5-point response scale of each item representing an appropriately sequenced set of skill points. Results indicated that nearly half of the COR-2 items had reversed or poorly spaced thresholds, suggesting potential problems with these items' functioning. Practice or Policy: Specific implications of the findings for the further development of the COR-2 in terms of its constructs and items as well as general implications for early childhood assessment are discussed.

Notes

¹The eight assessments selected were the (a) Creative Curriculum Developmental Assessment; (b) Galileo Preschool Assessment Scales; (c) HighScope Child Observation Record; (d) Learning Accomplishment Profile–Third Edition; (e) Learning Accomplishment Profile–Diagnostic; (f) Learning Accomplishment Profile–Diagnostic, Spanish Edition; (g) Mullen Scales of Early Learning; and (h) Work Sampling System (Halle et al., 2011).

Note. In the unstandardized model, one indicator per factor was fixed to 1 to define the factor metric (Brown, 2006). Categories and item titles are from HighScope (2010).

Note. Categories are from HighScope (2010).

Note. Salient pattern loadings (≥.40) are in bold. Item R loaded saliently on the Social Engagement and Scientific Process Skills factors and was removed from both. Item titles are from HighScope (2010).

Note. Loadings were transformed with the Schmid–Leiman procedure. Item titles are from HighScope (2010). Est = factor loading; %v = percent variance explained; h ² = communality.

Note. SE is the standard error for the slope parameter estimate; δij is the threshold parameter, with each item having four thresholds (j), or one fewer than the number of skill points (j = m − 1). Item titles are from HighScope (2010). GPCM = generalized partial credit model.

^a Item thresholds are disordered.