Informed Theory and Measurement Equivalence in Child Development Research

Abstract

Abstract— The role of theory informed by the nature of the population being studied is often overlooked in considering the equivalence of measures across developmental changes, across groups that differ in developmental status, or across groups that differ in participant characteristics. The purpose of this article is to discuss the “best practice” approaches for evaluating measurement equivalence based on such informed theory. This article discusses the role of informed theory in the consideration of various types of measurement equivalence in developmental research, as well as equivalence of translated items and measures.

Measuring the development of inhibitory control: The challenge of heterotypic continuity

“…Thus, longitudinal factorial invariance in the situation where different but overlapping measures have been used has the potential to greatly inform our understanding of heterotypic continuity in development (Nesselroade & Estabrook, 2009). However, longitudinal factorial invariance is not sufficient by itself to claim that equivalent measures reflect the same underlying construct and, as a result, is no substitute for construct validity invariance (Knight & Zerr, 2010), as we describe later. We recommend testing longitudinal factorial invariance whether the same or different manifest variables are used, because evidence of failed longitudinal factorial invariance can be informative of developmental change in the phenomena of interest (Edwards & Wirth, 2009; Widaman et al, 2010).…”

“…Alternatively, one can gradually relax parameter constraints for longitudinal factorial invariance to determine with which measures or at what ages longitudinal factorial invariance becomes established (and by extension, where it fails). It will then be up to the researcher to determine, ideally based on additional research, whether the failed longitudinal factorial invariance resulted from change in the manifestation of the construct, changes in the functioning of the measures, or some combination of the two (Knight & Zerr, 2010), because there is no statistical method for choosing between these alternatives. Unfortunately, there are few empirical guidelines for determining the severity of failed factorial invariance (i.e., whether it owes to changes in the functioning of a measure; Knight & Zerr, 2010), although some have proposed using effect sizes or other estimates (Millsap, 2010).…”

“…It will then be up to the researcher to determine, ideally based on additional research, whether the failed longitudinal factorial invariance resulted from change in the manifestation of the construct, changes in the functioning of the measures, or some combination of the two (Knight & Zerr, 2010), because there is no statistical method for choosing between these alternatives. Unfortunately, there are few empirical guidelines for determining the severity of failed factorial invariance (i.e., whether it owes to changes in the functioning of a measure; Knight & Zerr, 2010), although some have proposed using effect sizes or other estimates (Millsap, 2010). If the factorial invariance violations show important differences in the meaning of measures, say at a few of the ages in a series of measures of a construct, it may be necessary to exclude the measures at those ages from the analysis.…”

“…If the factorial invariance violations show important differences in the meaning of measures, say at a few of the ages in a series of measures of a construct, it may be necessary to exclude the measures at those ages from the analysis. Alternatively, one might substitute other measures reflecting the latent construct at those ages, as long as the set of measures under consideration remains a representative selection for the construct (Knight & Zerr, 2010). …”

The stability of elementary school contexts from kindergarten to third grade

“…Without it, we run the risk of mistaking differences in measurement over time for changes in key constructs (Bollen, 1989). There have been recent calls for greater attention to measurement equivalence within the fields of developmental science and human ecology (Knight & Zerr, 2010). However, to our knowledge, no study to date has examined whether the measurement of school context changes over time.…”