Modeling Theories and Theorizing Models: an Attempted Replication of Miller-Cotto & Byrnes’ (2019) Comparison of Working Memory Models Using ECLS-K Data

Longitudinal reciprocal relations among reading, executive function, and social-emotional skills: Maybe not for all.

“…There was no longitudinal reciprocal relation between reading and EF in the general population; this is in line with Feldon and Litson (2021), who did not find a longitudinal reciprocal relation between working memory and reading in a general population sample based on LGM-SR modeling. We expected this finding, which can be explained by sample effects proposed by the academic-cognition bidirectional theory (Peng & Kievit, 2020).…”

“…Specifically, Willoughby et al (2019) revealed small cross-lagged effects between reading and working memory (βs = .010–.230) and between reading and switching (βs = .040–.160). While Feldon and Litson (2021) found no significant cross-lagged effects between working memory and reading, indicating that such relations may not have existed between the two skills in their sample. It remains unclear whether reading, EF, and SEL may have longitudinal reciprocal relations in elementary school students from Grades 2 to 5, and if so, to what degree.…”

Early experiences and school readiness: A within and between exploration of the Opportunity Propensity Model

“…It is likely that we obtained different results than prior work because of the statistical analyses used. Indeed, Feldon and Liston (2021) arrived at different conclusions than Miller-Cotto and Byrnes (2020) due to the choice of statistical model. Our current work underscores the need to attend to the statistical modeling technique used with longitudinal data while also incorporating explanatory factors that relate to trait-level differences over time.…”

“…More recent work has examined whether propensity factors serve as mediators of children's development (Byrnes et al, 2018;Byrnes, Wang, & Miller-Cotto, 2019). The current study adds to the literature in the following ways: by using random intercept cross lagged panel models which can adequately separate out the within and between person effects, which prior cross lagged analyses cannot successfully do and actually biases our estimates, potentially demonstrating significant effects when they do not actually exist (Feldon & Liston, 2021). Further, the RI-CLPM is informative from a practical standpoint in the following ways: rather than focusing only on the mean level effects across all individuals, it offers a test of whether the OPM variables can explain stable, between-person differences over time while simultaneously estimating withinperson reciprocity in outcome variables.…”

“…Willoughby et al's (2019) cross-lagged effects are measurements of within-person stability given their use of linear growth curve with structured residuals; Miller-Cotto and Byrnes (2019) used a cross-lagged panel model that may conflate within-and between-person differences (Berry & Willoughby, 2017). In addition, Feldon and Liston (2021) provide a reanalysis of the Miller-Cotto and Byrnes study and, using models to account for within-and between-person variability, were unable to replicate the original Miller-Cotto and Byrnes findings of a transactional relation between math and WM. Specifically, a bivariate linear growth model of math and WM with structured residual cross-lags between math and WM fit well, but the cross-lags between math and WM were not consistently positive.…”