Figure 6.1 provides a summary of frequencies for the first group of Likert-type questions in the academic survey, presented at question 6. These questions related to general aspects of group work and were measured on a 7-point Likert scale (where 1 = very strongly disagree, to 7 = very strongly agree).
Given the small sample size (23), the large number of variables, and the 7 levels of measurement, the measurement categories were initially combined into a trichotomy comprising: total disagreed, undecided, and total agreed. This helps to make the presentation more readable and any patterns more obvious (de Vaus, 2002). Overall, academics agreed with most of the statements posed at question 6. Notably, the greatest level of agreement related to the belief that group work helps students engage in their learning (Q6d), χ2 (3, N = 23) = 12.652, p = .005, and that group work is an important aspect of university learning (Q6F), χ2 (4, N = 23) = 15.043, p = .005, (91.3% and 91.4% respectively). Group work’s provision of a ‘real-world experience’ for students (Q6c) (82.6%), was the third most frequently agreed upon statement in question 6, although it was not statistically significant (p = .052). Likewise, there was general disagreement that
Figure 6.1 Frequency of academic responses to general group work questions
0 10 20 30 40 50 60 70 80 90 100 Perc ent ag e o f a ca dem ics
General questions about group work
Total Disagreed Undecided Total Agreed
139 group work was a hindrance to a student’s ability to think and act independently (Q6j) (82.6%), but again this was not statistically significant (p = .088). The implications for staff workloads were not as clearly delineated, with 52.2% disagreeing overall, that group work increases workload for staff, while 43.5% agreed that it added to workloads.
Measures of central tendency, based on the original 7-point Likert scale measures, supported the combined frequencies of the trichotomy. Table 6.3 presents the frequency table for central tendency and variation in the question 6 items. The mode represents the most frequent response to each question, however since the Likert-type questions are ordinal, de Vaus (2002) argues that the median is the preferred measure of central tendency, because the mode is less stable and dependent on how categories are combined or collapsed, although notably, the wider the range, the less adequate is the median measure (de Vaus, 2002). In this instance, the mid-point (median) in the range of responses (1-7), is for the most part, the same as the most frequent response (mode). This has occurred despite the responses being dispersed across the full range, from those who strongly disagreed or very strongly disagreed, to those who very strongly agreed, for all except Q6d (helps students engage) and Q6c (a real-world experience), where there is less variation across the sample. The variation ratio provides an appropriate snapshot of the degree of difference that exists across the 7-point Likert-type question items in question 6. As shown in Table 6.3, only one item, ‘6D. Helps students engage in learning’ (Q6d), has a variation ratio below 50% (44%). This means that for all other items more than half the participants’ responses were not within the modal category indicated as being the most frequent. de Vaus (2002, p. 223) warned that ‘the more variation there is
Table 6.3 Academics’ responses to general aspects of group work
N Median Mode Min. Max. Var. Ratio 6A. Increases workload for staff 23 3.0000 3.00 1.00 7.00 0.65 6B. Helps students to master course material 23 5.0000 5.00 2.00 7.00 0.65 6C. Provides students with a real-world experience 23 6.0000 6.00 3.00 7.00 0.57 6D. Helps students engage in their learning 23 6.0000 6.00 3.00 7.00 0.44 6E. Stimulates students to work beyond minimum
requirements 23 5.0000 5.00 2.00 7.00 0.74
6F. Is an important aspect of university learning 23 6.0000 6.00 2.00 7.00 0.57 6G. Is an effective way of dealing with assessing
large classes 22 5.0000 6.00 2.00 7.00 0.70
6H. Is generally perceived negatively by students 22 5.0000 5.00 1.00 7.00 0.65 6I. Forms a planned and integral part of the whole
course in which teamwork skills are developed incrementally
21 4.0000 5.00 1.00 7.00 0.74 6J. Hinders students’ ability to think and act
140 in a sample the less well the averages summarises the sample’. These preliminary results suggest, that for this sample of academics, they are consistent in the direction of their responses, that is they agree (and disagree) in general, with the broad statements about group work, however the extent or degree of their perception differs. To further analyse these differences, bivariate and multivariate tests were conducted.
6.3.1.1 Bivariate analysis
A cross-tabulation table and a Pearson’s chi-square test of contingencies (α = .05), with exact statistics, to account for the small data set, and the sparsely populated cross tabulation cells (Mehta & Patel, 2012), was examined to further analyse the question 6 items. They were first assessed for potential relatedness to the demographic characteristics of the respondents, and secondly to the extent to which academics used group work in their teaching of accounting (Q7).
The initial outcome of the cross-tabulation and chi-square tests showed that for each of the question 6 Likert-type items, the expected frequency assumption, that stipulates ‘no more than 20% of the expected cell frequencies should be lower than five’ (Allen & Bennett, 2012, p. 229), was violated. This was a limitation associated with the small sample size. However, Mehta and Patel (2012, p. 16) argue that the ‘at least 5’ rule is ‘unnecessarily conservative’ for cross-tabulations and non-parametric tests. They propose that in these circumstances the Exact test or Monte Carlo two-sided p value, should be used, as they ‘provide a powerful means for obtaining accurate results when your data set is small…or the data fail to meet any of the underlying assumptions necessary for reliable results using the standard asymptotic method’ (Mehta & Patel, 2012, p. 1). The justification is that, by default, SPSS calculates statistics using the asymptotic method, and therefore assumes the data are of a sufficiently large sample size to fit a particular distribution. With small sample sizes, it is therefore ‘preferable to calculate a significance level based on the exact distribution of the test statistic’ (Mehta & Patel, 2012, p. 1).
The subsequent cross-tabulation, using the exact test option in SPSS, found a marginally significant association between the academics’ years of teaching experience in universities, and their perception that group work helps students engage in learning, χ2 (33, N = 23) = 47.622, p = .048, exact p = .025. Specifically, 100% of those with 10 to 15 years’ experience (30% of the sample group) either strongly or very strongly agreed that
141 group work aids engagement in learning. Those with twice as long a service record (30 years) disagreed with the statement. Although these individuals represented only 9% of the sample population, they were also the only academics to disagree with the statement. Assessing the influence of dichotomous demographic variables, gender, and teaching qualifications, respectively, a Mann-Whitney U test indicated that female academics were significantly more likely to rank group work as an important aspect of university learning (Mean Rank = 15.78, n = 9), than their male counterparts (Mean Rank = 9.57, n = 14), U = 29.00, z = -2.290, p = .022, two tailed, with an exact p = 0.21 (2- tailed). For the question 6 items, there was no indication that having a formal teaching qualification influenced responses. Interestingly, however, the extent to which group work was used (question 7) was significantly higher for academics without a formal teaching qualification (Mean Rank = 14.31, n = 16), than those with a qualification (Mean Rank = 6.71, n = 9), U = 19.00, z = -2.670, p = .008, two tailed, and an exact p = .007.
Table 6.4 shows that only 17.4% of the respondents rarely or only occasionally used group work in their teaching. The majority (82.6%) suggested that they often or always, used group work, which should be expected given the purposive sample of accounting academics currently using group work in their teaching. Not surprisingly this result was statistically significant, χ2 (3, N = 23) = 10.217, p = .017, exact p = .015.
However, 94% of academics without a teaching qualification were included in that statistically significant majority, compared with 57% of those who indicated they held a teaching qualification. For the question 6 items, there were no significant relationships between the ten general perceptions of group work listed in question 6 and the extent to which group work was used.