Abstract:

Psychometric networks are a recently proposed computational-statistical tool for the analysis of multivariate data in psychology. The networks consist of graphical representations of the relationship between multiple variables measured in a set of individuals. The nodes represent the variables and the interconnections (edges) between nodes are association parameters between variables, for example, partial correlation coefficients. Networks allow the study of psychological problems at the systems level, and the organization of its elements, to achieve the purposes of describing, predicting, and controlling the psychological phenomena of interest.

To date, few or no studies have applied the network methodology in the context of educational evaluation. This could be very relevant for studying the psychosocial determinants of students' academic performance. Outstanding determinants are socio-emotional skills, which are non-cognitive factors that allow students to regulate their emotional states, relate effectively to their peers, and make responsible decisions. Which have a positive impact on the prevention of violence or addictions, and also promote their entry into the labor markets.

In this study, free data from the 2017 national assessment in Mexico called PLANEA was used, which included 117,700 students in the last grade of upper secondary level from 2,319 schools in Mexico. The evaluation consisted of multiple educational, sociodemographic, and psychological measures, within the psychological ones included measurements of the socioemotional skills of grit (8 items), social awareness (8 items), emotional self-regulation (5 items), and decision-making (15 items). We built a network of partial correlations between the 37 items of the psychological measures and a measure (1 item) of the academic grades obtained during upper secondary education. The observations of 112,566 students who did not have missing data in any of the variables of interest were used.

The results revealed a network of homogeneous and strongly associated nodes that belong to the constructs of social awareness and self-regulation. The nodes that made up the decision-making and perseverance constructs were more heterogeneous since some of their components were more associated with nodes of other dimensions. The network made it possible to identify that the items that were most related to academic performance are those of the grit construct, for example, the more a student considers that he/she works hard, he/she also obtains better grades; or, the more difficulty a student has in paying attention, the worse his/her grades were. Likewise, the topology of the network identified that the most important node, in terms of its strength and number of connections, was that of the decision-making construct that says "I try to be clear about my objectives before deciding".

The use of the psychometric network methodology can positively contribute to the exploration of the elements within the socio-emotional skills that may be more relevant for academic performance. This could suggest potential intervention targets that positively impact the delivery of effective and comprehensive education.

Keywords:

Network psychometrics, partial correlations, soft skills, Mexico, multivariate statistics.