Principal component analysis
The PCA was carried out using SPSS. Varimax was used as a type of extraction for facilitating the interpretability of the factors because "orthogonal rotation methods assume that the factors in the analysis are not correlated" (Gorsuch, 1983, pp. 203–204, cited in Dean, 2009 p. 21). One solution that emerged from the PCA became the basis of the cluster analysis. The first step of the PCA was to check whether factor analysis could be applied to the data set obtained and to see the suitability of the 80 non-nominal items, on which we can potentially apply a factorial analysis, using the Kaiser-Meyer Olkin (KMO) and Bartlett tests. High values in KMO (about 1.0) generally indicate that a factorial analysis can be useful with the data, as well as the value close to 0.000 in the Bartlett sphericity test. Test results confirm that a factorial analysis was useful with our variables, reflecting a score of 0.894 which is considered remarkable in KMO with a value close to 0.000 of Bartlett’s sphericity test sufficient.
The PCA extracted 19 core component solutions spread over 4 large blocks that we show in Table 1. Eight of the main components made up the basis of the cluster analysis that we describe below for being closely related to the research questions.
The second step towards profiling respondents consisted of a cluster analysis to identify students based on a conglomeration (segmentation) technique. According to Allen (2017), "the choice of input variables becomes important both to provide the basis of similarity within a grouping and to differentiate the differences between groupings" (p.143).The final solution main components that have emerged from the PCA have been used as a basis for creating students profiles, using a clustering technique (segmentation), depending on the frequency of activities, resources and interactions they use for keeping up-to-date and their use of social media.
Finally, eight component variables were identified as suitable for responding to the purpose of the investigation, shown in Table 1. The first PCA block represents two components that group the use of distinctive social media platforms. Below, the PCA solutions that have been used in our process of clustering are presented, together with the associated questionnaire variables.
The next step of the cluster analysis consisted in establishing the total number of clusters of respondents. For this purpose, the methodology of previous studies that have used similar analytical procedures through SPSS was followed (Kahan et al., 2017; Guitert et al., 2018; Poellhuber et al., 2019). We have performed a hierarchical cluster analysis to define the possible cluster range. Once the analysis of hierarchical clusters was performed, we observed by means of a dendrogram as the number of clusters was mainly adjusted to 4, 5 and 6 possible or appropriate solutions. Secondly, we have followed the K-means procedure to form and set the definitive number of clusters. To determine the optimal solution, the analytical procedure tested 4, 5 and 6 categories, comparing the quality of the different models and the meanings of the profiles produced. A classification model was sought in which the profiles would be qualitatively and significantly different, while trying to preserve the quality of the final solution. Finally, the optimal number of clusters was set to 5. Table 2 shows the comparative characterization of the profiles of respondents identified through the cluster analysis based on the components that have emerged from the PCA (Table 3).
Figure 2 illustrates the different scores of the 5 different students’ profiles that have emerged from the K-means method solutions. The figure shows the profiles of respondents based on the resources, activities, and relations they use to learn and their use of social media platforms.
Cluster solutions have been named through an interpretive process and then compared to each other to maximize differences and similarities. Below we present a description of the learning profile solutions in light of the component variables.
Learner profile 1: “Versatile” students learning through multiple and varied contexts, with strong support from Wikipedia, Blogs and YouTube (n = 108). The first cluster represents 19.1% of the total sample and is characterized by significantly high scores on most variables, the highest among all clusters, with no low scores. The learning ecologies of this group are rich, distributed, and conscious. Regarding the activities in which they participate, we find that this group has the highest score in formal face-to-face actions (courses and events), followed by remarkable scores in digital and informal actions, which can be used as complementary support for formal learning. Regarding resources, they use more formally guided online resources than formal print resources and UOC material, with little difference between the two. Regarding interactions, we found slightly higher scores with groups outside the academic field. Finally, their use of social media is very unequal: they make great use of platforms with static core and organizational content features (Wikipedia, Blogs and YouTube) and a more discreet use of social network platforms.
Learner profile 2: “Dependent” students with preference for academically guided learning resources (n = 107). The second cluster represents 19% of the total sample and is characterized by significantly high scores on variables in place with academically guided learning resources and a moderate score on the rest of the update resources, while especially low on those that have to do with professional contexts. Cluster 2 learning ecologies are strongly dependent on academic indicators. In relation to the learner activity component, we find a low participation in formal face-to-face actions and moderate participation in digital actions that are less formally guided. Regarding the resource’s component, we found a high score on formal resources on paper and material of the UOC, so it seems to rely basically on formal content offered from the academic field, which is supported by the other variable with high score, such as interacting with academic groups. In contrast, online resources and information source platforms occupy a neutral position and social network oriented platforms have the lowest negative score among the 5 clusters.
Learner profile 3: “Self-guided” students with preference for informal and digital learning resources and with support for interactive social network platforms (n = 113). The third cluster represents 20% of the sample and is characterized by high scores in survey variables that have to do with more informal and digital resources, actions and interaction and a moderate and low score on those more formal and analog resources. This group is in clear contrast from the previous cluster 2 profile. Cluster 3 learning ecologies are self-guided and happen through digital contexts. In relation to the learner activity component, we observed a high score on mostly digital and informal actions and a low score on formal face-to-face actions (courses and events). Regarding the resource’s component, the trend is repeated, showing a high score (the highest of all clusters) on less formally guided digital resources and a low score on formal resources on paper and UOC material. Regarding the learner interaction component, the cluster again displays a higher score in interactions with groups outside the academic realm than within it. Regarding the use of social media, unlike the previous cluster, we see a preference for social and interactive network platforms (Twitter, LinkedIn, Instagram, and Facebook) and discreet use of Wikipedia, YouTube, and blogs.
Learner profile 4: "Analog" students with preference for physical and formal resources, with some support for interactive social media platforms (n = 133). The fourth cluster represents 23.6% of the total sample, the highest percentage among all identified groups, and is characterized by moderate and high scores on variables that have to do with analog resources, physical resources, and interactions outside of an academic context and social network-type platforms and low scores on all other variables. Cluster 4 learning ecologies take place through face-to-face actions and slightly formal mixed resources, establishing connections on social networking platforms. Regarding the learner activity component, this cluster participates more in face-to-face and formal actions such as courses and events, than in digital and informal activities. As far as the resource component is concerned, we find that they use formal resources on paper and UOC material, making little use of online resources. Regarding the interaction component, learners interact more with groups outside the academic context than within it. However, the cluster makes considerable use of interactive social network platforms.
Learner profile 5: “Detached" students academically, with low use of resources for learning and occasional social media (n = 103). The fifth cluster represents 18.3% of the total sample and is characterized by moderate to low scores on the 8 questionnaire variables. In general, cluster 2 learning ecologies have a low degree of use of actions, resources, and connections for professional development, as this cluster demonstrated the lowest scores of all groups. To some extent, it reveals findings that are opposite to the cluster 1 profile. In relation to the activity component, we find that among the low values, this group participates more in formal face-to-face actions than in more informal digital actions. Regarding the resource component, it is the opposite, this learner profile makes slightly higher use of digital resources that are less formally guided. Also, in line with low engagement we find little interaction with teachers and students, and, to a lesser extent but also low, with professionals and friends. Finally, students in this cluster make sporadic use of social media.