**3. Results**

*3.1. Descriptive Analysis*

Regarding the descriptive analyses (see Table 3), the LESI items could be grouped around four values:



**Table 3.** Descriptive results of LELI scale.

#### *3.2. Reliability and Sample Adequacy*

The sample for internal consistency analysis was 796. The Cronbach's Alpha index was used to check the level of reliability. Table 4 shows the results of Cronbach's α for the LESI scale, obtaining appropriate results (0.809).


#### *3.3. Exploratory Factor Analysis*

Exploratory Factor Analysis (EFA) was applied using the Principal Components method and Varimax rotation for the LESI scale. Previously, the Kaiser–Meyer–Olkin Sample Adequacy Measure (KMO = 0.767) was performed (see Table 5), which showed a high partial correlation coefficient. This result evidenced that the variance was not caused by underlying factors. Subsequently, the Bartlett's test of Sphericity was 4416.101 (p153 < 0.001), which demonstrated that there is no relationship between the test items.

**Table 5.** Kaiser–Meyer–Olkin (KMO) and Bartlett's test.


The analysis provides a structure of five factors that explain 60.158% of the total variance. During the process, coefficients lower than 0.3 were suppressed. Figure 2 shows a sedimentation graph of the factorial structure of five factors on the abscissa axis and the eigenvalues on the ordinate. The factors with high variances are located in the first five components, evidenced by a steep slope. After the fifth component, there is an evident change in slope, correlating with a weaker interpretation of the construct.

**Figure 2.** Sedimentation graph of the factor components.

The variables were grouped as follows (see Table 6):



**Table 6.** Exploratory Factor Analysis results.

#### *3.4. Multiple Linear Regression Analysis*

In order to verify the prediction assumption, a multiple linear regression analysis of the sample was applied using the Stepwise method. For this, the 18 LESI variables were included. As a result, a general model of seven variables was established (see Table 7): Learning space attachment (v5), Wall color satisfaction degree (v10), Importance of interactions with students from other courses (v18), Acoustics satisfaction level (v11), Ventilation satisfaction degree (v8), Importance of professor–student interactions (v15) and Practice classroom favors teacher–student interactions (v3). The model explains 7.6% of the academic outcome (GPA), establishing direct and inverse relationships. In addition, the Durbin-Watson statistic is 1.519, which fulfills the assumption of residual independence (Table 7).

**Table 7.** Multiple linear regression results: LESI variables on GPA.


The Pearson correlations of the seven-variable model are provided in Table 7. Positive Beta values, indicating a direct relationship with the academic outcome variable, are associated with five variables: Learning space attachment (v5), Wall color satisfaction degree (v10), Acoustics satisfaction level (v11), Ventilation satisfaction degree (v8), Importance of professor–student interactions (v15). Negative Beta values, indicating an inverse relationship with the academic outcome, as provided by the remaining variables: Importance of interactions with students from other courses (v18) and Practice classroom favors teacher–student interactions (v3).

In addition, it must be verified that perfect multicollinearity does not exist, to validate the model; for this, the variance inflation test was applied. Table 8 provides the VIF values that are close to one, indicating no collinearity problems or correlation between the input variables. Furthermore, the tolerance values are also close to one, so the other independent variables do not explain any of them in particular.


**Table 8.** Standardized coefficients and collinearity statistics: LESI variables on GPA.

Another of the assumptions to be checked is linearity; Figure 3 displays the values that predict our estimation with respect to the values of the regression residuals. The result confirms the assumption of homoscedasticity, since the variance is practically homogeneous for the entire range of values. This figure also demonstrates compliance with the principle of linearity, since there is no non-linear pattern in the data cloud.

**Figure 3.** Cloud points of standardized predicted values vs. standardized residuals.

The last check requires that the distribution of the residuals follow a pattern close to normality. The P-P plot verifies compliance since, in general, the factors are close to or above the line (see Figure 4).

Once the validity of the model is verified, it is necessary to analyze the ANOVA results (see Table 9). This provides an F statistic value and an associated probability value, as well as sums of squares, degrees of freedom, and mean squares. A probability value less than 0.05 indicates that the model is consistent, thus allowing us to explain the relationship between the input and output variables.

**Figure 4.** Normal P-P plot of Regression Standardized Residual. Dependent variable: GPA.

**Table 9.** ANOVA results.


Finally, Figure 5 shows the partial regression graphs of each variable in the model, where the line is the equation obtained from the linear regression analysis.

The multiple linear regression analysis performed on LESI demonstrated the existence of a relationship with academic outcome. The coefficient of determination was 0.076, while the mean square error was 0.9540.

**Figure 5.** Partial regression graphs: LESI variables on GPA. (**a**) Learning space attachment; (**b**) Wall color satisfaction degree; (**c**) Importance of interactions with students from other courses; (**d**) Acoustics satisfaction level; (**e**) Ventilation satisfaction degree; (**f**) Importance of professor–student interactions; (**g**) Practice classroom favors teacher–student interactions.

#### **4. Discussion and Conclusions**

Sustainable building design is one of the priorities for the preservation of resources and energy. In recent decades, research based on post-occupation evaluation studies of indoor environment quality indicators have become more and more common. Factors such as geographic, cultural and climatological diversity have confirmed that it is not possible to develop a single model, but rather it is essential to disseminate research to diagnose reality. In addition, it is important to consider the functions, uses and habits in order to determine a sustainable design. In the case of educational buildings, in addition to IEQs, teaching methods should be taken into consideration. At present, several university models involve active teaching, which promotes the participation and interaction of students as a basis for learning. This research aimed to delve into three constructs—the peer effect, place attachment and IEQ satisfaction—and their relationship with academic outcomes.

For this, the Learning Environment and Social Interaction scale was designed and validated, which evidenced a structure of five factors: classroom design as a facilitator of social interactions, working design satisfaction, the learning value of social interaction, classroom environmental satisfaction and place attachment. Regarding the first factor, the literature shows that flexible designs allow teachers to promote greater learning and better adaptability to active methodologies, which also allows an improvement in the flow of social interactions between students [29,30]. The IEQ satisfaction has was divided into two factors: workspace design and classroom environmental satisfaction. The literature does not really show this separation, but there is agreemen<sup>t</sup> on the variables related to the activity performed in the space and those related to the environment [38,69]. Regarding the learning value of social interaction, positive relationships between classmates have shown benefits for academic performance and motivation [34,35]. Finally, place attachment has been confirmed as a factor in itself; as the basis of the link between the person and the place [60], it not only improves the well-being of the users of the building but also favors inclusion and interaction between people [57].

The regression results indicate that place attachment is the LESI variable that explains academic performance in the sample to the greatest extent. This contribution supports previous studies that demonstrated that place attachment had a higher value than other common IEQs, such as lighting, regarding the development of academic performance [67]. Regarding IEQ satisfaction, wall color, acoustics and ventilation also evidenced a direct relationship with the academic outcome, in line with evidence on satisfaction with the learning environment [44,45] and quality of environment [46].

Two variables of the learning value of social interaction showed an inverse and a direct relationship with the learning outcome: the importance of interactions with students from other courses and the importance of professor–student interactions. These findings indicate that greater interaction with the teacher may be related to a better understanding of the objectives or content of the subject and consequently of the academic outcome, while greater interaction with students from other courses leads to lower solvency of the course. Although it seems a consistent result, it is common for students to lean on peers from other courses, as they create bonds of friendship beyond academic assignments. However, this scenario would require a larger study to determine whether the support of outsiders is correlated with fewer interactions in the classroom itself. Previous research has verified the negative effects of a greater number of classmates and interactions and the positive effects of interactions in a smaller group [26], which could identified as the academic group within the classroom.

Finally, only one of the variables of classroom design as a facilitator of social interaction showed an inverse relationship with the academic outcome: the practice classroom favors teacher–student interactions. This result shows an apparent contradiction, in line with the literature [23,24]. Previously, it came to light that those students who assign higher value to teacher–student interactions obtain better GPAs, and those users who perceive that the classroom favors these interactions obtain worse results. The research does not allow for determination of the reason for this difference, but it seems logical that those who value interaction for their learning use it for academic purposes, also known as "learning interaction". Meanwhile, those who indicate that the design favors interaction would use it for social purposes.

This leads to the conclusion that the social dimension of the physical space contributes to the explanation of a student's academic result. In addition, the research contributes to the design and validation of LESI, the scale that those responsible for higher education centers can apply to diagnose their particular scenario and consequently manage the pertinent modifications.

This research focused on quantitative methods, due to its exploratory nature. A multi-method approach would be beneficial to complement the theory on learning space satisfaction and social interaction in higher education. Furthermore, more research is necessary both in higher education and at other educational levels, so that the high costs of building can be justified by substantial data on sustainable architecture in terms of purpose adequacy. Likewise, research on each factor, allowing an in-depth understanding of the particular complexity of each variable, is required.

**Author Contributions:** Conceptualization, V.L.-C.; methodology, V.L.-C.; software, V.L.-P.; validation, V.L.-C. and V.L.-P.; formal analysis, V.L.-C. and V.L.-P.; writing—original draft preparation, V.L.-C. and V.L.-P.; writing—review and editing, V.L.-C. and V.L.-P. Both authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study. The participation was optional and entailed the treatment, analysis and publication of findings.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.
