Mediational effect of prior preparation on performance differences of students underrepresented in physics

This study examined the mediation and moderation of membership in a demographic group underrepresented in physics classes on course outcomes measured by course grades and Force and Motion Conceptual Evaluation (FMCE) post-test scores. The study used a large dataset ($N=4490$) of course grades, SAT and ACT mathematics scores (ACTM), and matched FMCE pretest and post-test scores to investigate differences by gender, underrepresented ethnic or racial minority (UERM) status, and status as a first-generation college student (FGCS). For UERM and FGCS students, ACTM and pretest scores significantly mediated the relation of membership in the demographic group and both course grade and post-test score. Differences between minority and majority members of these groups were largely removed by controlling for ACTM and pretest scores. The overwhelming majority of the effect acted through ACTM for course grade (60% and 45%, respectively), while more of the effect acted through pretest score for the post-test (36% and 48%, respectively). As such, for these groups prior preparation measures predict physics outcomes (course grades or post-test scores) differently. The mediational relations for gender were dramatically different. No mediation was detected for the relation of gender to course grade because no significant difference in course grade existed. Sixty percent of the effect of gender on post-test score was not explained by either ACTM or pretest score; pretest score accounted for 38% of the effect. As such, the majority of the difference in post-test scores between men and women was not explained by either ACTM or pretest scores. Significant moderation was also detected showing that the relation of these variables was not consistent for members of all demographic groups.

Figure 1

Mediation process.

Figure 2

Path diagrams of the relation of gender, UERM status, and FGCS status, ACTM, pretest scores, post-test scores, and course grades. A path model shows the results of multiple regression analyses. The dependent variable of each regression is the node at the head of the path. The nodes at the tail of the path represent the independent variables. The number on each path is the regression coefficient for the independent variable. For the path with two numbers reported, the number outside of the parenthesis is the regression coefficient using only the demographic variable; the number inside the parenthesis is the regression coefficient with the demographic variable and the other variables included in the model. Note superscripts denote (*) $p<0.1$, (a) $p<0.05$, (b) $p<0.01$, and (c) $p<0.001$.

Figure 3

Moderated mediation examining the effect of gender on grades (a) and post-test score (b). Gender is abbreviated Gen; Pretest as Pre. Note superscripts denote (*) $p<0.1$, (a) $p<0.05$, (b) $p<0.01$, and (c) $p<0.001$.

Figure 4

Moderated mediation examining the effect of underrepresented minority status on grades (a) and post-test score (b). Pretest is abbreviated as Pre. Note superscripts denote (*) $p<0.1$, (a) $p<0.05$, (b) $p<0.01$, and (c) $p<0.001$.

Figure 5

Moderated mediation examining the effect of first-generation status on grades (a) and post-test score (b). Pretest is abbreviated as Pre. Note superscripts denote (*) $p<0.1$, (a) $p<0.05$, (b) $p<0.01$, and (c) $p<0.001$.