Capturing the Relationships Between Achievement Goals and Outcomes: Latent Profile Analysis Versus Multiple Regression

• In Event: Considering the Whole Student: The Use of Person-Centered Approaches to Characterize and Study Motivation

Abstract

Achievement goal theory represents one of the dominant frameworks for understanding students' achievement motivation (Elliot, 2005). The 2-factor model distinguishing between mastery and performance goals initially put forth by achievement goal theorists (e.g., Dweck, 1986) was eventually expanded to distinguish between approach and avoidance strivings. This occurred first for performance goals (e.g., Elliot & Harackiewicz, 1996) and then for mastery goals (Elliot & McGregor, 2001). As we continue to refine the types of achievement goals that students pursue, it is essential that we evaluate how different goals work in combination to promote or undermine students' motivation.

One method researchers have adopted to examine the relationship of goals with other variables is multiple regression (MR; e.g., Elliot & Church, 1997; Harackiewicz et al., 1997; Kaplan & Midgley, 1997; Middleton & Midgley, 1997). However, the use of MR becomes increasingly complicated as the conceptualization of achievement goals becomes more complex. Models with multiple main effects and interactions can be difficult to interpret and are limited in their ability to capture nonlinear relationships between variables. In contrast, person-centered analyses provide an alternative approach, and in previous research we explored the merits of using cluster analysis (Pastor, Barron, Davis, & Miller, 2004) and latent profile analysis (LPA; Pastor, Barron, Davis, & Miller, 2007) over MR. In the current study, we investigate single-step LPA, a relatively new technique proposed by Bauer and Shanahan (2007), which differs from other applications of LPA in that an outcome variable is used along with achievement goal variables when creating profiles.

To illustrate the benefits of single-step LPA, we compared the results of MR and single-step LPA by examining the relationship between achievement goals and GPA for 3,345 college students who completed a modified version of Elliot and McGregor’s (2001) achievement goal questionnaire. Although the results for all four goals will be reported, we only consider the results from mastery-approach (MAP) and performance-approach (PAP) here. The MR results indicate a significant interaction between MAP and PAP, with MAP having a stronger influence on GPA at lower levels of PAP. When the predicted values of GPA are plotted for various levels of MAP and PAP, the linear relationships of the goals with GPA are evident. The single-step LPA also captures this interaction, but its plot of predicted values differs in that the relationships of the goal orientations with GPA are nonlinear. Comparing these two approaches illustrates: (a) how both are able to capture interactions, (b) how MR allows only linear relationships whereas single-step LPA allows the relationships to be of any form.

To acquire the single-step LPA results described above, MAP, PAP and GPA were used as latent class indicators. The resulting 5-class solution illustrates the utility of including outcomes along with goal orientations in a single analysis. For example, two profiles emerged with moderately high levels of MAP and PAP, but very different average GPA values. In the full paper, the achievement goal profiles emerging from our analyses are described and compared to those obtained by other researchers.

Authors

• Dena A. Pastor, James Madison University

• Kenn E. Barron, James Madison University