Predicting Student Success in a Psychological Statistics Course Emphasizing Collaborative Learning

Research on Collaborative Learning

The benefits of collaborative learning approaches have been well documented. Studies have found these techniques to be associated with more positive course evaluations (Barchard & Pace, 2010; Shibley & Zimarro, 2002), greater success in achieving course goals (Barchard & Pace, 2010; Comfort, 2011; Nevin et al., 1994), more positive student relationships (Nevin et al., 1994), and student perceptions of enjoyment and educational benefit (Hulbert-Williams, 2010). Two meta-analyses (Johnson et al., 2000; Springer et al., 1999) found that a range of cooperative learning approaches have a significant positive effect on student achievement and that cooperative approaches fared better than either competitive or individualistic approaches. There is even evidence to suggest that performance of individual group members on course outcomes can be affected by satisfaction with or endorsement of a group work approach (Besser, 1995; Shaw, Duffy, & Stark, 2000).

Although there is ample evidence that collaborative learning in a general sense can increase student learning (Smith, Sheppard, Johnson, & Johnson, 2005), a recent meta-analysis of best practices that enhance collaborative learning (based on studies published in Teaching of Psychology) is of particular relevance. Tomcho and Foels (2012) found that group activities that contained high levels of participant interdependence were associated with greater learning outcomes, a finding underscored by other researchers as well (Michaelsen & Sweet, 2008). This aligns with the structure of the team activities in our statistics courses, which incentivize students to check each other’s work and collaborate on a series of group questions after they complete individual problem sets.

There may also be limitations to the use of collaborative learning in higher education settings. Delucchi (2007) noted that the empirical evidence for the effectiveness of such approaches in statistics courses is mixed at best. Fransen, Kirchner, and Erkens (2011) concluded that, in many cases, collaborative teams in higher education focus more on aspects of performance and less on team aspects. In developing a measure for assessing peer-based collaborative learning, Pazos, Micari, and Light (2010) pointed out that problem-solving approach and group interaction style are keys to successful collaborative learning, suggesting that not all groups will be equally successful. Remedios, Clarke, and Hawthorne (2008) described the particular challenges of “silent” participants in collaborative learning contexts, and others have noted the problem of “freeloaders” in group contexts (Shibley & Zimarro, 2002). All of this evidence suggests that collaborative learning methods must be implemented with care and that the success of such approaches is contingent on a number of factors. As Slavin (1996) has argued, although there is a growing consensus about the benefits of collaborative learning, there is still much to be understood about the circumstance under which cooperative learning is most effective.

Course Structure and Collaborative Tasks

In our department, statistics for psychology is an entry-level requirement for majors and minors which serves hundreds of students each year. As is the case in undergraduate psychology curricula across the country, a basic understanding of statistics is an essential tool for students’ successful completion of course requirements, is beneficial to their critical thinking in general, and is necessary to their understanding of the scientific paradigm as psychology majors. One of the challenges to teaching the course is its size, with approximately 100 students enrolling in each term. This large class size results in diminished opportunities for faculty members to work closely with students who therefore miss opportunities to identify and modify misconceptions. Over the last several years, our department has developed a collaborative system for fostering learning by providing peer support and practice in application of course material with the notion that such group learning benefits the students by allowing them to support and structure each other’s learning (Gredler & Shields, 2008).

Based upon a pretest of statistics knowledge, students in the course are divided into five groups representing each quintile (80th to 100th percentile, 60th to 79th percentile, etc.) and then randomly assigned to 1 of 20 groups consisting of 5 mixed-ability students. At various intervals during the term, groups work together on class problem sets and conceptual questions, with all members receiving the same scores. This comprises roughly 40% of the total class time. This approach adopts a mixture of problem-based learning and the cognitive apprenticeship model where more skilled students assist those with less expertise (described in depth by Enkenberg, 2001).

As an example, one of these tasks supports the students’ learning of the concept of correlation. First, each team member completes an individual problem set that involves computation of a correlation coefficient using the appropriate formulas and interpretation of the computed results. Second, team members exchange their work with fellow teammates and check each other’s individual work for errors. Finally, the team reunites to discuss a series of questions that can only be answered by looking at all the individual components in concert. So, in the correlation example, students list which sets of computations led to a statistically significant correlation. Students also reflect together on more conceptual issues; in the case of correlation, they describe a study methodology that would allow for a causal claim about the relations between the variables for which they previously computed a correlation. Other key course concepts that were applied include z-scores, z-tests and the hypothesis-testing model, power, effect size, and confidence intervals, single-sample t-tests, independent and dependent sample t-tests, one-way analysis of variance, and chi-square analyses.

Study Goals and Hypotheses

Although course evaluations indicate that students predominantly like this system and anecdotal evidence suggests that it helps to facilitate learning, a better understanding of individual differences that predict success in a course such as this would be helpful. As noted earlier, there is ample evidence that collaborative learning can contribute positively to student learning and can be effective in a college classroom context, but we also wondered about the utilization of such approaches in a statistics course, where students would presumably enter the class with strong attitudes both toward the course material itself and toward the collaborative learning structure. Indeed, earlier work has shown that anxiety about course content can substantially affect students’ experiences in a class (Conners, McCown, & Roskos-Ewoldson, 1998; Gourgey, 1984; Zeidner, 1991). It is possible that students who are more open to a collaborative approach could have lower levels of anxiety (Hugh-Jones & Madill, 2008). Prior work that has specifically examined the impact of collaborative group work on performance in statistics courses has found inconsistent evidence (Delucchi, 2007), suggesting that the benefits of collaborative approaches may be tempered by individual differences in variables such as student attitudes.

Participants

Data were collected from approximately 200 students across two classes, although the final analysis was conducted on the 166 participants who provided complete data. The two sections were taught by two different instructors utilizing the same approach.

Materials

Participants completed two measures, namely, a Feelings Toward Group Work Questionnaire and an additional survey designed by the researchers that assessed interest in statistics, relevance of statistics, and anxiety toward learning statistics.

Procedure

On the first day of class, we informed students about the study and asked if they would consent to participating. Students then completed the Feelings Toward Group Work Questionnaire, the attitudes toward statistics survey, and the pretest of statistical knowledge. The instructor then randomly assigned students to mixed ability teams representing each quintile of the distribution of scores. Teams met on approximately 40% of class meeting days.