Which measures work for indirect family-school engagement in early childhood? A comparison of predictive validities
*Names in bold indicate Presenter
The current study builds upon past research by (1) identifying several techniques for measuring home-ECE partnerships and (2) comparing the concurrent and predictive validity of the various measurement types.
I use data from the Fragile Families and Child Wellbeing study, which follows nearly 5,000 children born in large U.S. cities between 1998 and 2000 and their families. The data includes an oversampling of families at high risk of breaking up or entering poverty, which allows these analyses to give important insight on those populations. This study focuses on a subset of 571 children who were observed in their home and ECE settings at age three. In both environments, observers responded to identical questions on the responsivity and harshness of the caregiver using the HOME Inventory (Bradley & Caldwell, 1984). I use these parallel measures to compare these two environments. I also use data from the year five and year nine follow-up waves for longitudinal child outcomes.
The measurement techniques that I compare are first, a simple median split for family and ECE contexts that divides children into four categories for high-high, high-low, low-high, and low-low based on their scores across contexts. Second, match scores indicate whether home and ECE environments received the same score on specific items of responsivity and harshness. Finally, previous research has used cluster analytic techniques to identify naturally occurring clusters of responsivity and harshness in this dataset. I use these previously-identified profiles. The assumptions and implications of each measurement technique are discussed.
I then compare these three measurement techniques using multiple regression methods to predict child outcomes. Preliminary analyses have shown that each of these measurement techniques are predictive of later child outcomes, even after controlling for family and ECE characteristics, but that the patterns of prediction differ depending on the measurement used.