Cooperative Learning in Introductory Statistics: A Mixed-Methods Assessment of Students' Perceptions, Performance, and Learning

• In Event: Innovation in Introductory Statistics Instruction: Student Attitudes, Motivations, and Peer Support as Key to Learning

Abstract

Objectives
In an introductory statistics classroom, some students struggle to understand key concepts and may become intimidated by peers who seem more knowledgeable. The use of active, discussion-based, peer-to-peer learning methods including cooperative learning (CL) is highly recommended in statistics instruction (ASA, 2016) and may afford opportunities to advance all students’ understanding by exploiting the diversity of prior knowledge in the class. We experimentally compare homogeneous grouping and heterogeneous grouping as two alternative ways of organizing CL in introductory statistics and examine their process and outcome in terms of students’ performance, learning, and perception of group dynamics.

Framework
CL has been advocated as an effective way of organizing instruction to promote students’ engagement and achievement (see Johnson et al., 2006; Slavin, 1983; 1996). The hypothesized mechanisms include motivational, affective, and cognitive benefits of peer support (De Lisi, 2002; Johnson et al., 2015). However, questions regarding CL’s efficacy remain, including: 1) whether CL is more effective when the ability levels of group members are homogeneous (Piagetian perspective) or heterogeneous (Vygotskyan perspective), 2) whether CL facilitates students’ conceptual learning or merely boosts their performance, and 3) what social-emotional processes underlie CL and whether they depend on group composition. Such information could elucidate the mechanisms by which CL shapes the learning process.

Methods
We recruited 142 university students from five introductory statistics classes. We first categorized students as relatively lower (LPK; n = 88) or higher (HPK; n = 54) prior knowledge on the basis of their prior course-taking and preliminary statistical knowledge. Within each class, we randomly assigned LPK students to homogeneous (LPK-HOM) or heterogeneous (LPK-HET) groups of three to four, the latter containing at least one HPK student. Throughout the term, students completed three CL activities reviewing key statistical concepts recently taught. In each exercise, students first independently completed a pre-test, then collaborated on a post-test containing identical items (performance gains), and then independently completed a test containing new but relevant items (learning gains). An end-of-term survey investigated student experience during CL.

Results
Hierarchical linear models revealed that students in both homogeneous and heterogeneous groups showed considerable gains in performance and learning. Within the heterogeneous groups, LPK-HETs had higher gains in performance than HPKs; yet LPKs and HPKs did not differ in their eventual learning gains on average. Qualitative analyses of students’ free responses revealed additional benefits of CL including camaraderie, emotional support, and sharing of alternative perspectives. Students were largely in favor of CL, although LPK-HETs reported having relatively less to contribute compared to LPK-HOMs and HPKs.

Significance
This study addresses the theoretical controversy about how best to implement CL in introductory statistics classes. Contrary to the theoretical hypothesis that LPK students would benefit if participating in CL with HPK students rather than with other LPK students, we found that group composition did not make a difference in learning. Even though the social dynamics may differ between heterogeneous and homogeneous groups, CL activities appeared to promote engagement and conceptual understanding regardless of a student’s prior knowledge or group composition

Authors

• Almaz Mesghina, Northwestern University

• Guanglei Hong, University of Chicago

• Adelle Durrell