Profiling Change: Methods for Comparing Patterns in Mathematics Expectancies and Values Across Time

• In Event: 61.015 - Embracing and Modeling the Complex Dynamics of Motivation and Engagement: Contextual, Temporal, Dynamic, and Systematic

Abstract

Aim and Theoretical Framework
This study examines mathematics motivation in elementary school through the lens of Eccles’ Expectancy-Value Theory in which academic decisions are informed by the combined influence of whether a student expects to be successful at and values an activity (Wigfield & Eccles, 2000). Generally, these constructs are viewed separately, in a variable-centered approach; however, this may ignore how they interact within individuals as part of a holistic self. We take a person-centered approach (Magnusson, 1998; e.g., Corpus & Wormington, 2014) to consider patterns of expectancies and values and how these patterns change throughout the school year.
Method and Data Analysis
Students in our sample are 9,749 fourth and fifth graders in four districts across the United States (Table 1). Questions addressed expectancy, usefulness, and importance of mathematics to students now and in the future and were part of a larger survey embedded in a digital mathematics environment that was given in the fall, winter, and spring of the 2017-2018 school year.
We used two different statistical methods to examine mathematics motivation: (1) latent profile analysis, examining each time point separately, with latent transition analysis to understand changes in profiles between (LPA/LTA), and (2) Latent profile analysis including information from all three time-points together, which incorporates student change in expectancies and values in the formation of the profiles (LPA-3). Analysis of single-time point profiles and change between is more common in person-centered approaches (e.g., Nurmi & Aunola, 2005); our large sample size provided power to also investigate change as part of a profile itself.
Results and Significance
We discuss fourth graders here; fifth graders will be included in the presentation. Profiles for both LPA/LTA and LPA-3 are provided in Figures 1-4; transition probabilities are provided in Table 2-3. Both methods revealed interesting distinctions between future and current usefulness and importance in cluster formation. Clusters were formed among students who held lower expectancies and current values, yet viewed mathematics as useful and important for their future at each time point and in the combined LPA-3. The LPA-3 showed overall trends not picked up within the LPA/LTA. For example, LPA/LTA could describe which students moved between high and low profiles, but didn’t pick up on the movement within the profiles themselves. As examples from the LPA-3: a profile that started out in fall near but not at the lowest and declined across time relative to others; a profile that experienced a decline in winter motivation but recovered in spring; and a profile that jumped in motivation from fall to winter and sustained at these higher levels.
Although a person-centered approach fits within a dynamic view of development (Bergman, 2001), as typically analyzed, this approach still looks at patterns as states, examining change between. Using our LPA-3 model, we show that such analyses may miss important variance in profiles of motivation. Our presentation will explore this further by looking at which students get placed in each profile within both models and which method is more predictive of academic outcomes.

Authors

• Sarah Marina Karamarkovich, North Carolina State University

• Teomara Rutherford, University of Delaware