Capturing family-school partnership constructs over time: Creating developmental measurement models

Abstract

Longitudinal research methods have become increasingly popular with researchers interested in understanding how and why outcomes change over time. Recent developments in statistical methodology and the availability of software with which to conduct such research have made longitudinal methods more accessible. These include latent growth models, which allow researchers in the area of family-school partnerships to investigate issues such as how parental involvement in students’ schoolwork changes over time and how changes in parental involvement relate to changes in students’ achievement levels. The estimation of longitudinal models has traditionally been based on use of the same items at each time point. However, this may pose a problem because items that are developmentally appropriate for younger students may not be appropriate for older students. In this chapter we propose and illustrate developmental measurement models that are appropriate for measuring student outcomes over time, but that do not necessarily include the same items at each age or grade level. These models explicitly allow for items to be dropped from or added to the scale in order to maintain developmental appropriateness, while maintaining a common set of items. Inclusion of the common items provides a basis on which the scores for each age group to be linked or equated such that they are on the same scale. Thus, developmental measurement models make it possible to conduct longitudinal research using scales that are appropriate to each age group.