PARA ESTUDIANTES UNIVERSITARIOS STRUCTURAL VALIDITY OF THE LEARNING SELF-REGULATION QUESTIONNAIRE FOR UNIVERSITY STUDENTS

The study’s aim was to obtain evidence of validity of the Learning Self-Regulation Questionnaire (LSRQ) internal structure and to verify metric invariance compared with a previous research study. The participants were 237 university students from the first three academic semesters of a private university in three Peruvian cities (two in the North of Peru, and one in Lima). The analysis was performed by a semi-confirmatory factor analysis, specifying as comparison matrix: a) the configuration derived from a previous study, and b) the free estimation loadings factors. The results indicate that two dimensions represent the instrument structure satisfactorily; but the metric invariance compared to a previous study was not satisfactory. The re-specification of the model, by removing two items with factorial complexity problems and the free estimation of the items, was successful. These results are discussed so as to the interpretation of their scores and the lack of metric invariance.


Massification of higher education in different
contexts during the last decades (Rama, 2008) has lead school students, who have not yet achieved maturity, to decide what program o study, as part of their tertiary education. Thus, adolescents between 16 and 17 years old, when being admitted to the university, shall adapt to the new educational demands (Chau y Saravia, 2015). The university freshman usually discovers that he or she must modify several organizational aspects, such as study methods, time devoted to academic activities and the effort to obtain success and guarantee a professional future (Gutiérrez et al., 2010).
For this to be possible, their learning process will need to be self-regulated, based on their individual liberty and responsibility.
Students who self-regulate tend to be more proactive and make a bigger effort to learn (Rosario et al., 2014). They become cognitively, emotionally and behaviorally conscious (Rosario, et al., 2009); they are aware of their capabilities and limitations (Rosario et al., 2014); their behavior when approaching education is based on their personal goals and strategies (Rosario et al., 2014;Rosario et al., 2009); they contemplate about their progress (Rosario et al., 2014), and improve on how they measure their learning achievements (Lopez, Hederich-Martinez and Camargo, 2012;Hernandez Pina, De Fonseca and De Tejada, 2010;Rosario et al., 2012). Also, self-regulated learning benefits from the self-worth perception one has, as well as from the developed self-efficacy that can come in handy in the future (Rosario, et al., 2012), and the teaching strategies carried out by educators (Chaves, Trujillo and Lopez, 2015;Gaeta, 2014;Santelices, Williams, Soto and Dougnac, 2014) in learning environments, both individually and collectively (Jarvela, 2015).
One of the many viewpoints there is on self-regulation in learning is the one proposed by Deci and Ryan (2002). According to them, in their presented model about selfdetermination, people's conduct is based on their self-management and empowerment; in other words, their own motivation will lead them to commit to learning. Ryan and Deci (2000) also emphasize the importance of inherent resources and innate psychological needs of autonomy, competitiveness and relationships in order to produce selfmotivation and mental health. Something that is also examined is how certain social environments work against the development of the aforementioned learning tendencies. mediante un análisis factorial semiconfirmatorio, especificando como matriz de comparación (a) la configuración derivada de un estudio previo, y (b) la estimación libre de las cargas factoriales.
Los resultados indican que dos dimensiones representan satisfactoriamente la estructura del instrumento; pero la invarianza métrica respecto a un estudio previo no fue satisfactoria. La reespecificación del modelo, mediante la eliminación de dos ítems con problemas de complejidad factorial y la estimación libre de los ítems, obtuvieron resultados satisfactorios. Se discuten estos resultados en el marco de la interpretación de sus puntajes y la falta de invarianza métrica. As pointed out by Moreno and Martinez (2006), self-regulated learning synthesizes theories about cognitive evolution, organic integration, and orientation in purpose and in basic needs. It also incorporates theories about: innate motivation, self-control, personal competence, motivation from certain tasks, positive feedback, and the perception of effectiveness, selforientation and self-government.
Based on their theory, Williams and Deci (1996) (Hooper, Coughlan and Mullen, 2008;Hu and Bentler, 1999). Apparently, some structural aspects were not resolved, such as the possible factorial complexity of some items, or the existing relationship between an item and two or more irrelevant latent variables Grimaldo, 2010, 2011). Lack of specificity of this characteristic can have an impact on decreasing fit indexes because the source of variability of the items is not modeled (Brown, 2006

Procedure
It's an instrumental-type research (Montero and Leon, 2002). The approximate time was 15 minutes.
As for the analysis, it consisted of examining the internal structure of the LSRQ through a semi-confirmatory procedure. This method prevents the identification of the dimensional structure from being oriented from the data, as is the case in the exploratory factor analysis (Nunnally and Bernstein, 1995).
It also prevents the modeling from being highly restrictive, as is the case in a confirmatory factor analysis/structural equation modeling (Nunnally and Bernstein, 1995 consisted of seeing if the items' factorial loads remained the same; something also known as "metric invariance" (Elosua, 2005). The specific procedure was done through the rotation of the factorial solution of the present data towards the reported configuration in Table   2

Preliminary Analysis
The items showed a clear pattern in the answer trend, since those representing the autonomous behavior were more frequent, while the control ones were fewer and slightly more dispersed (see Table 1). On the other hand, the non-rotated factorial solution seems to show a pattern which is different in its convergent factorial loads between the items and their constructs.

Correlation among Factors. Using
Matos' (2009) specification, the interfactorial correlation was moderate and negative (see Table 2), while in the partial specification and the modified model correlation was zero (see Table 3) and practically the same as observed Given the content of the analysis, which was based on a polychoric matrix, the previous estimates are attenuated (Elosua and Zumbo, 2008), because that they were calculated based on a different matrix. Based on the matrix of polychoric correlations, the ordinal internal consistency (α) (α o ; Elosua and Zumbo, 2008) was calculated through an ad hoc program (Domínguez, 2012). It was found that for F1 (α o = .91) and F2 (α o = .87) reliability was higher so they can be considered more appropriate estimates. The calculated α o was done without the two items that were removed (5 and 13) from F2. In this analysis same context, the estimated reliability from a congeneric model (ω) was inferior in relation to α o (see Table 3).  An aspect that seems to have been influenced by the lack of replicability is the correlation between constructs. When adjusting the data to Matos' (2009) structural parameters, the association between factors was negative.
Some aspects were not evaluated and the same are part of the limitations in the current study. In first place, the invariance of all the items' parameters (intercepts and residuals) was not evaluated, which would have allowed the comparison between groups in various aspects, such as the means and variances between them (Elosua, 2005). This limitation is applied to the comparison between our data and Matos' (2009) study, and between the relevant groups in the current investigation, for example, men and women, or between semesters. The second unestimated aspect indicates that the sample size can have an effect on obtaining high stability and a higher sampling error within the gathered statistics, compared to larger samples. This means a bigger sample size is required in order to obtain more representative conclusions. Also, the similarity between the factorial loads within each factor was not verified. This is known as tau-equivalent (Meyer, 2010) and it helps in the interpretation of the factor and to back up the use of coefficient α as the appropriate estimator for internal consistency reliability.
Lastly, the interpretation of both constructs in the current sample indicates that these can coexist when the subject faces academic situations. The linear correlation between both, equivalent to zero in this study, suggests that control and autonomy do not have to be opposite behaviors, they can actually interact and adapt to each other in order to obtain successful learning results. Nevertheless, this last point needs to be looked into and research dedicated to it is guaranteed.