*2.2. Analysis*

The analysis process commenced with the exploration of demographical information, followed by an analysis of the collaboration, learning, and usefulness (CLU) characteristics (Table 2). Here means and standard deviations were identified for describing data distribution variations [53]. The empirical relationship between the CLU variables were tested by performing two bivariate regression analyses, where the first one tested the relationship between collaboration (C) and learning (L), and the second tested the relationship between learning (L) and usefulness (U). Demographical data was collected exclusively for description purposes, thus not integrated into the regression analysis. The working assumption was that focusing on collaboration enhancing elements during a collaboration exercise leads to collaboration learning, which again leads to usefulness [50]. In this first test, collaboration was defined as an independent variable relative to learning (dependent), while in the second test, learning was the independent variable to usefulness (dependent). Pearson correlation coe fficients (Pearson's r) were calculated to measure the linear dependence between the variables. Coe fficients of determination (r2) were calculated to determine what proportions of the variance in the dependent variables could be considered predictable from the independent variables [54]. To test the di fference in parameters between the observed results and the stated hypotheses, standard errors and F-values were calculated. Further, the significance level (*p*-value) was calculated to determine the probability of rejecting the null hypothesis. Statistical significance was set to *p* < 0.05, and all tests were two-tailed [49]. In the next step, the relevant predictor and criterion variables that were found significant were tested in multiple regression analysis [48]. Here collaboration and learning were used as integrated independent variables relative to usefulness (dependent variable). The standardized coe fficient values (beta), and the di fferences between them, were calculated both for the bivariate and multiple regression analyses to analyze the strength of the e ffects of each CLU variable. Towards the end, *p*-values were calculated together with the performance of a Shapiro-Wilk test [55]. The last was done to ensure that the CLU variables met normal assumptions for regression analysis.



Note: n = 71, sign = *p* < 0.05.
