Measurement Invariance of Working Memory in Early Childhood: A Registered Report

• In Event: Developing Assessments for All Early Learners: Towards a More Equitable Understanding of Young Children’s Development

Abstract

Executive function is one of the most investigated constructs in both cognitive science and in education given its high correlation with numerous academic and life outcomes. Differences in executive function skills between children from higher socioeconomic and lower socioeconomic homes, as well as children from different racial/ethnic backgrounds, with children from underresourced and otherwise minoritized communities demonstrating poorer performance relative to their peers with more resources. However, a number of factors might lead to children from different communities approaching widely used tasks in different ways—particularly at the start of kindergarten when prior experiences are particularly diverse—yielding systematic differences in performance that may be attributable to something other than actual differences in executive function skill.

The goal of the present study is to determine whether the Numbers Reversed task—arguably the most widely used measure of working memory in the educational assessment landscape in the US—is invariant across racial/ethnic groups. That is, is this standardized assessment functioning similarly for all students, or is the assessment systematically biased toward one group or another? In the case that the task is not invariant across racial/ethnic groups, comparisons of means cannot be made across racial/ethnic groups and conclusions of prior work that has compared EF across groups may be invalid.

We first identified our analytic sample from the larger Early Child Longitudinal Study–Kindergarten Cohort of 2010-11, a large-scale, prospective, longitudinal study with participants recruited to be nationally representative. Due to our focus on measurement invariance across racial/ethnic groups, we chose to focus on the most widely compared groups (who were also most widely recruited within this sample). The resulting analytic sample comprised participants identified as non-Hispanic/Latino Asian (n=1360), Black or African American (n=2135), and White (n=7985), and Hispanic/Latino (n=4280).

To test measurement invariance, we will first fit a single-factor model to dichotomous item-level data from the Numbers Reversed task with all participants. We will then run multi-group models testing for measurement invariance at configural, metric, and scalar levels. As this is a registered report, analyses have not yet been performed but results will be available at time of presentation. Implications for past and prior research, as well as considerations for equitable approaches to measurement will be discussed.

Authors

• Dana Miller-Cotto, Kent State University

• Andrew Ribner, University of Pittsburgh
  Presenting Author

• Sammy Ahmed, University of Rhode Island

• Alexa Ellis, University of Alabama, Tuscaloosa