Attitude toward scientific research and academic self-efficacy as statistical predictors of self-perceived research skills in university students in Peru

Introduction: Scientific research has transformed universities; however, in countries such as Peru, gaps in knowledge production persist, making it necessary to strengthen research training and promote evidence-based practices. This study aimed to determine whether attitude toward scientific research and academic self-efficacy statistically predict self-perceived research skills among students at a private university in Peru.

Methodology: The study employed a quantitative approach with an explanatory–associative scope and a non-experimental, cross-sectional design. A non-probability convenience sampling method was used, and data from 650 university students (52.8% female and 47.2% male) were analyzed. The variables were measured using the Attitude toward Scientific Research Scale (EACIN-R), the Academic Self-Efficacy Scale, and a scale of self-perceived research skills. Confirmatory factor analyses conducted with the present sample supported the factorial structure of the instruments, while reliability evidence reported in the original validation studies indicated adequate internal consistency. Measurement models were treated as multidimensional. The use of observed variables was theoretically justified.

Results: Regression analysis indicated that attitude toward research ( β = 0.145; p < 0.001) and academic self-efficacy ( β = 0.678; p < 0.001) significantly and statistically predict self-perceived research skills. Positive correlations were identified between attitude toward research and academic self-efficacy ( r = 0.431; p < 0.001), between attitude toward research and self-perceived research skills ( r = 0.437; p < 0.001), and between academic self-efficacy and self-perceived research skills ( r = 0.740; p < 0.001). The correlations of 0.431 and 0.437 were moderate, while the correlation of 0.740 was strong and reflects perceived rather than actual research ability.

Conclusion: Attitude toward scientific research and academic self-efficacy are significant statistical predictors of self-perceived research skills, with academic self-efficacy emerging as the strongest predictor. These findings highlight the relevance of both psychological factors and methodological training in university research education. Results should be interpreted considering the self-report nature of the instruments and the limitations inherent to a cross-sectional, non-experimental design, which does not allow causal inferences.

Scientific research has been associated with significant transformations in university academic environments, promoting evidence-based practices that strengthen knowledge innovation and improve educational quality (Devos et al., 2019). In countries such as Peru, however, structural gaps in knowledge production persist, alongside urgent social demands that require solutions grounded in rigorous research practices (Loayza-Rivas, 2021). In this context, self-perceived research skills have become essential competencies for planning, designing, executing, and communicating research processes aimed at understanding and addressing social problems. These skills involve students' subjective evaluation of their cognitive, procedural, and attitudinal abilities to engage in scientific inquiry (Domínguez and Fernández, 2019; Ipanaqué-Zapata et al., 2023). Strengthening reading, technological, and digital capabilities is also crucial for accessing, selecting, and using reliable information through the effective management of databases, scientific journals, and virtual libraries (Nuñez-Pacheco et al., 2023).

Globally, scientific production in the social sciences shows a marked geographic concentration, with the United States leading with 21.97% of publications, followed by the United Kingdom (5.77%) and China (5.52%). Peru ranks 28th worldwide with only 0.98% of global production (SJR, 2024a). Within Latin America, Brazil leads with 31.64% of publications, followed by Chile (14.76%) and Peru (14.01%) (SJR, 2024b). Although these figures reflect progress, they also reveal persistent gaps in the consolidation of research competencies in the region, consistent with studies documenting inequalities in global scientific production (De-Moya-Anegón et al., 2018; Kwiek, 2021). These structural gaps are mirrored in university research training, where pedagogical and psychological factors are associated with the development of scientific competencies.

In professional training contexts, academic self-efficacy plays a central role. It is defined as the personal perception of one's ability to organize and execute actions aimed at achieving academic goals (Gutiérrez-Torres, 2018). Empirical evidence consistently shows that self-efficacy is a robust statistical predictor of academic performance (Honicke and Broadbent, 2016; Schunk and DiBenedetto, 2021). In Peru, most students report average levels of persistence and academic self-efficacy, and strengthening these attributes is recommended due to their association with perceived research performance (Domínguez and Fernández, 2019).

Despite advances in research education, self-perceived research skills remain a challenge for university students. Studies report insufficient methodological preparation (Rojas et al., 2020) and emphasize that teachers' research competencies are strongly associated with the development of such skills (Pastor et al., 2020). Among Latin American students enrolled in thesis courses, difficulties are most evident in formulating research problems, objectives, and hypotheses (51.63%), as well as in analyzing and processing data using statistical techniques (49.37%) (Ipanaqué-Zapata et al., 2023). These challenges are associated with academic overload, burnout, limited scientific reading, and weaknesses in theoretical and practical training (Chara and Olórtegui, 2018). International literature also shows that attitudes toward research and formative experiences are associated with motivation and engagement in scientific activities (Brazendale et al., 2024; Lopatto, 2007; Seymour et al., 2004). Recent studies on research competencies further highlight their relevance for sustainable academic development (García-González et al., 2022).

From a broad conceptual perspective in higher education, scholars emphasize that the development of transversal competencies associated with academic and scientific training—such as critical analysis, systems thinking, methodological rigor, and complex problem-solving—extends beyond the strict boundaries of research. In this regard, the entrepreneurship education literature posits that this set of foundational cognitive skills also serves as an essential framework for navigating innovation- and entrepreneurship-oriented learning environments (Bacigalupo et al., 2016).

In alignment with this perspective, Jardim, 2021 highlights that competencies such as autonomy, task persistence in complex environments, reflective self-assessment, and structured communication of results act as key catalysts for fostering an entrepreneurial mindset and innovative capabilities among students. Consequently, although research training and entrepreneurship education are often treated as distinct domains, a functional convergence exists wherein research skills enhance students' ability to identify opportunities, manage uncertainty, and formulate innovative solutions in dynamic contexts.

Although scientific advances have expanded the understanding of research-related constructs, a conceptual gap persists regarding the joint analysis of attitude toward research, academic self-efficacy, and self-perceived research skills. Attitude toward research refers to students' beliefs, evaluations, and affective predispositions toward the research process. Academic self-efficacy, grounded in Bandura, 1977 sociocognitive theory, reflects general beliefs about one's capacity to organize and execute actions necessary to achieve academic goals. In contrast, research self-efficacy refers specifically to confidence in performing research-related tasks, such as formulating research questions, collecting data, or interpreting findings. Likewise, self-perceived research skills represent students' subjective assessment of their competence in carrying out research activities, rather than objectively measured research performance or research competence. Although these constructs share a common component related to perceived competence, they differ in their level of specificity and should not be considered interchangeable. Students who value research positively and trust in their academic abilities are more likely to engage in research-related activities and report stronger self-perceived research skills. However, empirical evidence in Peru remains limited regarding the interaction and statistical associations among these constructs.

Given that these skills are measured through self-report instruments, they may be influenced by subjective biases, potentially inflating associations among variables. Additionally, the instruments used in this study present multidimensional structures, which must be considered when analyzing their measurement properties. The use of observed variables is theoretically justified to examine direct associations among constructs. Consistent with the original validation studies, global scores were employed to represent the broader constructs measured by each instrument.In this context, the general objective of the present study was to determine whether attitude toward scientific research and academic self-efficacy statistically predict self-perceived research skills in undergraduate students at the Peruvian Union University, Juliaca campus, providing updated empirical evidence to strengthen research training in Peruvian higher education.

From a theoretical perspective, students who hold favorable attitudes toward scientific research are more likely to engage in research-related activities, value the research process, and persist in the development of investigative tasks. These dispositions may contribute to stronger self-perceptions of research skills. Likewise, sociocognitive theory proposes that individuals with higher levels of academic self-efficacy tend to demonstrate greater confidence when facing academic challenges, invest more effort in learning activities, and persist when encountering difficulties, which may be associated with stronger self-perceived research skills. Furthermore, positive attitudes toward research and academic self-efficacy are expected to be associated because students who perceive themselves as capable of meeting academic demands are more likely to approach research activities with confidence and favorable expectations. Consequently, attitude toward research and academic self-efficacy may jointly contribute to explaining variability in self-perceived research skills.

Based on the theoretical and empirical evidence reviewed, attitude toward research and academic self-efficacy are expected to be statistically associated with and contribute to the prediction of students' self-perceived research skills.

• H1. Attitude toward scientific research and academic self-efficacy statistically predict self-perceived research skills in undergraduate university students.
• H1a. Attitude toward scientific research statistically predicts self-perceived research skills in university students.
• H1b. Academic self-efficacy statistically predicts self-perceived research skills in university students.
• H1c. Attitude toward scientific research is significantly associated with academic self-efficacy in university students.

Study focus, design and scope

The present study adopted a quantitative approach, as systematic procedures and numerical measurements were used to contrast the proposed hypotheses regarding attitude toward scientific research, academic self-efficacy, and self-perceived research skills, following the guidelines of empirical–analytical research (Creswell and Creswell, 2018). The design was non-experimental because the study variables were not manipulated, and cross-sectional because the information was collected at a single time point (Hernández and Mendoza, 2020).

The scope was explanatory-associative, as the relationships between variables were analyzed to examine the statistical contribution of attitude toward research and academic self-efficacy to self-perceived research skills through regression analysis (Ato et al., 2013). This scope is consistent with the use of regression models, which evaluate the relative statistical contribution of psychological variables to a specific criterion without establishing causality. Given the non-experimental, cross-sectional design, the results should be interpreted as evidence of statistical prediction and association rather than causal relationships.

Although the attitude toward research and self-perceived research skills instruments have multidimensional structures, both scales were originally developed to provide an overall score representing a broader construct. Consistent with the original validation studies and supported by the adequate fit of the measurement models in the present sample, global scores were used in the regression analyses to examine associations among the study variables.

Population and sample

The population consisted of 8,000 undergraduate students from the Peruvian Union University, Juliaca branch, enrolled in the 2025-II academic semester. The sample consisted of 650 students, selected through non-probability convenience sampling, a common criterion in educational studies whereby access to participants depended on institutional availability (Etikan, 2016).

Students from all faculties and academic cycles were included, provided they were enrolled and agreed to participate voluntarily. No additional exclusion criteria were applied. Prior to analysis, the database was reviewed to identify incomplete records or duplicate responses. Only fully completed questionnaires were retained for analysis. This procedure is consistent with non-experimental studies aimed at estimating statistical relationships rather than controlling experimental conditions (Creswell and Creswell, 2017; Etikan, 2016).

Instruments

Attitude toward scientific research

Attitude toward scientific research was measured using the Attitude Toward Scientific Research Scale (EACIN-R), validated in the Peruvian university population by Hidalgo et al., 2023. The instrument consists of 28 items distributed across three dimensions: Interest in research (9 items), Research vocation (12 items), and Evaluation of research (7 items). Responses were recorded on a 5-point Likert scale (“strongly disagree” to “strongly agree”).

The confirmatory factor analysis (CFA) reported by the authors showed adequate fit indices: χ2 = 690.222, GFI = 0.94, AGFI = 0.93, RMR = 0.05, NFI = 0.92, RFI = 0.915. Reliability was α = 0.898, demonstrating high internal consistency (Nunnally and Bernstein, 1994). These results provide solid evidence of validity and reliability for measuring attitudes toward scientific research.

Academic self-efficacy

Academic self-efficacy was measured using the scale developed by Domínguez, 2018, composed of 9 items with a 4-point Likert format (“never” to “always”). This instrument assesses academic self-efficacy as a global construct using a unidimensional structure supported by confirmatory factor analysis.

The CFA yielded a unidimensional model with adequate indices (χ2 = 64.687; RMSEA = .056; GFI = .969; CFI = .978), supporting the instrument's construct validity. Internal consistency was also adequate [α = .881, 95% CI (.864,.897)].

Self-perceived research skills

Self-perceived research skills were measured using the scale developed by Guerrero-Narbajo et al., 2023, composed of 17 items distributed across three dimensions: Thinking skills (7 items), Academic writing ability (7 items), and Information communication ability (3 items). This instrument evaluates self-perceived research skills, not objective performance, and therefore reflects self-assessments that may be influenced by subjective perception.

The CFA reported by the authors showed excellent fit indices (χ2/df = 1.237, CFI = 0.991, RMSEA = 0.019, SRMR = 0.041, TLI = 0.997). Reliability through total omega was Ω = 0.898, indicating robust internal consistency. The instrument is culturally appropriate for Peruvian university students and aligns with the construct of self-perceived research skills examined in this study.

Data collection procedure

Data collection was carried out through a virtual survey administered via Google Forms, which allowed efficient organization and storage of information in both remote and in-person environments (López and Fachelli, 2017).

Prior to data collection, authorization was obtained from the corresponding academic authorities and approval from the Ethics Committee of the Peruvian Union University. Participants provided informed consent, in which the nature of the study, voluntary participation, and data confidentiality were explained, following the ethical principles of the World Medical Association (2013). No personally identifiable information was collected, and the data were stored in a secure repository accessible only to the research team.

Given that all instruments were self-report measures, responses may be influenced by subjective biases inherent to self-assessment.

Data analysis

The data were processed using SPSS and AMOS. Descriptive statistics were computed: mean, median, skewness, and kurtosis (Tabachnick and Fidell, 2007).

Multiple linear regression analyses were conducted to determine the extent to which attitude toward research and academic self-efficacy statistically predict self-perceived research skills. Multivariate normality, absence of outliers, and independence of errors were evaluated following recommendations by Hair et al., 2019. Normality was verified using a normal P–P plot of residuals and skewness and kurtosis values. Multicollinearity was assessed using VIF and tolerance tests. The significance level adopted was p < .05.

Confirmatory factor analyses were conducted in AMOS using the Maximum Likelihood estimator, which has been widely used in the evaluation of Likert-type instruments and was considered appropriate given the sample size and distributional properties of the data. Following the scoring procedures proposed by the original authors, total scores were calculated and used in subsequent analyses because the purpose of the study was to examine relationships among broader constructs rather than among individual dimensions. The use of observed variables allowed the estimation of direct associations among constructs without modeling latent structures, consistent with the predictive scope of the study. In addition, linearity and homoscedasticity were examined through visual inspection of residual-versus-predicted-value plots.

Demographic characteristics

Table 1 describes the demographic characteristics of the participants. Most of the sample were women (52.8%), and nearly half were men (47.2%). Regarding age, 56% of the students were under 20 years old and 38% were between 21 and 25 years old, indicating a predominantly young population, which is typical of undergraduate university cohorts. With respect to place of origin, 52.2% of the participants came from the Andean region and 42% from the Amazon rainforest, reflecting strong representation from both areas. In terms of academic program, the largest group belonged to Administration (35.7%), followed by Environmental Engineering (14%). Finally, 46.2% of the students were in their ninth academic semester, indicating a greater representation of students in advanced stages of their university studies.

Table 1. Demographic characteristics of the participants.

These demographic characteristics contextualize the interpretation of self-perceived research skills, which reflect subjective evaluations rather than objective performance.

Measurement model fit indices

Table 2 presents the fit indices obtained from the confirmatory factor analyses (CFA) conducted with the present sample. In all cases, the χ2/df values were below 2, indicating an adequate level of model–data fit. The CFI and TLI values ranged from.923 to.978, reflecting good comparative fit and appropriate performance relative to more parsimonious models. RMSEA values ranged from.052 to.076, which fall within the acceptable range for models with moderately complex factorial structures. Likewise, SRMR values ranged between.025 and.035, indicating minimal discrepancies between the observed and estimated matrices.

Table 2. Model fit indices of the variables.

The multidimensional structure of the instruments was considered in the measurement models. Taken together, these results confirm that the measurement models for attitude toward research, academic self-efficacy, and self-perceived research skills exhibit a solid and psychometrically consistent factorial structure.

Given the adequate fit of the measurement models and considering that the original validation studies support the interpretation of overall scores, global scores were used in subsequent correlational and regression analyses.

Descriptive statistics

Table 3 presents the descriptive statistics for the study variables. Attitude toward research presented a mean score of 100.02 (SD = 23.04). Self-perceived research skills presented a mean score of 61.27 (SD = 11.18), whereas academic self-efficacy presented a mean score of 24.92 (SD = 5.63). Skewness and kurtosis values fell within the acceptable range (±1.5), indicating approximately symmetric and mesokurtic distributions. This stability in distributional shape supports the assumptions required for parametric statistical analyses, including the multiple regression model applied in the study.

Table 3. Descriptive analysis of the variables.

Regression assumptions and residual diagnostics

Table 4 presents the evaluation of regression assumptions and residual diagnostics. Tolerance values (.815 and.815) and VIF values (1.228 and 1.228) ruled out multicollinearity, indicating adequate independence between the predictors. The Durbin–Watson statistic (1.96) supported the assumption of residual independence. Residual normality was confirmed by the Shapiro–Wilk test (W = .990; p = .081), along with acceptable skewness (−.22) and kurtosis (.48). Influence indicators—Cook's distance (.006), leverage (.007), and standardized residuals (−2.51 to 2.68)—remained within recommended thresholds, suggesting the absence of influential cases. Taken together, these results demonstrate that the regression model satisfactorily meets the statistical assumptions required to ensure the validity and stability of the estimates.

Table 4. Evaluation of regression assumptions and residual diagnostics.

Predictor 1: Attitude toward research; Predictor 2: Academic self-efficacy.

Additionally, residual-versus-predicted-value plots were visually inspected and indicated no substantial deviations from the assumptions of linearity and homoscedasticity.

Outlier diagnostics

Table 5 presents the diagnosis of potential outlier cases. Standardized residuals ranged from −2.51 to 2.68, falling within the acceptable ±3 range and indicating no extreme deviations. Cook's distances were low (.003–.005), and leverage values ranged between.005 and.007, showing that none of the observations exerted disproportionate influence on the model. Taken together, these indicators confirm that no influential cases were detected and that all observations were retained for analysis.

Table 5. Diagnosis of potential outlier cases.

Correlation between variables

Table 6 shows positive and statistically significant correlations among all variables. Attitude toward research showed a moderate correlation with academic self-efficacy (r = 0.431; p < .001) and with self-perceived research skills (r = 0.437; p < .001). A correlation of considerable magnitude was observed between academic self-efficacy and self-perceived research skills (r = 0.740; p < .001), indicating that both variables share a substantial proportion of variance.

Table 6. Correlation analysis among the study variables.

This pattern is consistent with perceived, rather than actual, research ability and reflects the self-report nature of the instruments.

Multiple regression model

Table 7 presents the global indicators of the regression model. The adjusted coefficient of determination (R2 adjusted = 0.564) indicates that 56.4% of the variability in self-perceived research skills is jointly explained by attitude toward research and academic self-efficacy, indicating that the model accounts for a considerable proportion of variance. The ANOVA confirmed that the model was statistically significant (F = 421.016; p < .001), showing that the set of predictors is statistically associated with the criterion variable.

Table 7. Multiple correlation coefficient.

The regression model was estimated using observed variables, consistent with the analytical approach adopted in this study.

Reliability

Table 8 summarizes prior reliability evidence for the instruments used in this study. The original validation studies reported adequate internal consistency across all scales, with Cronbach's alpha coefficients ranging from.881 to.898, and a McDonald's omega coefficient of.898 for the Self-Perceived Research Skills Scale. These metrics reflect values from previous validation studies and were not re-estimated using the current sample of participants (N = 650).

Table 8. Reliability evidence of the instruments used in the study.

Reliability coefficients correspond to the original validation studies of the instruments. NR, not reported.

Regression coefficients

Table 9 presents the coefficients of the multiple regression model. Academic self-efficacy emerges as the strongest statistical predictor of self-perceived research skills (β = .678; p < .001), whereas attitude toward research also shows a significant association, although with a considerably smaller standardized coefficient (β = .145; p < .001). These results show that both variables are significantly associated with the criterion variable, with academic self-efficacy demonstrating the greatest explanatory contribution within the model. These associations should not be interpreted as evidence of causal influence.

Table 9. Multiple regression coefficients.

Regression model

The regression model explained 56.5% of the variance in self-perceived research skills (R² = 0.565). Figure 1 illustrates the statistical associations among the variables. Attitude toward research (β = 0.145) and academic self-efficacy (β = 0.678) were significantly associated with self-perceived research skills, while the predictors showed a moderate correlation between them (r = 0.431). The coefficients correspond to the standardized multiple regression model, and academic self-efficacy exhibited the strongest explanatory weight, underscoring its statistical relevance in students' perceptions of their research competencies. The arrows represent statistical associations, not causal pathways, given the cross-sectional design.

Interpretation of the main findings

Associations among the study variables

The results show a positive and significant relationship between attitude toward research and academic self-efficacy (r = 0.431; p < 0.001). Given the non-experimental, cross-sectional design of the study, this result should be interpreted as evidence of statistical association rather than causality. In a study with 500 university students, Amjad and Naseem, 2025 found a positive correlation between research attitude and research self-efficacy (r = 0.301; p < .01), although of a smaller magnitude than that observed in the present study. This difference suggests that, in the analyzed sample, a favorable attitude toward research is more closely linked to the perception of personal efficacy, which may be consistent with contexts where institutions emphasize research courses, training practices, or faculty mentoring. The instruments demonstrated adequate psychometric performance, supporting the robustness of the observed associations.

Current studies also report that both variables tend to increase simultaneously in contexts where research training is structured and experiential. In the quasi-experimental research of Falola et al., 2025, a three-week training program significantly increased the six dimensions of research self-efficacy and raised positive attitudes toward research from 74.6% to 81.0%. In contrast, Quintero et al., 2025 found that a brief, self-administered intervention produced only a marginal increase in favorable attitudes, without reducing perceived barriers. These results are consistent with the idea that attitude and self-efficacy tend to co-occur in training scenarios that provide real, guided, and systematically oriented experiences.

Similarly, the positive correlation between attitude toward research and self-perceived research skills (r = 0.437; p < 0.001) supports the notion that affective-cognitive dispositions are associated with students' self-perception of investigative competencies. This result is consistent with studies based on the theory of planned behavior, which propose that attitude is associated with investigative intentions and actions, favoring the development of scientific skills (Vera et al., 2026).

Recent, more analytically rigorous evidence supports this interpretation. Almarwani, 2025, using structural equation modeling, showed that the dimensions of attitude toward research were positively associated with perceived competence, emphasizing perceived usefulness as the strongest predictor (β = 0.68; p < .001). However, other studies caution that a positive attitude alone does not guarantee competent performance, as evidenced by Ipanaqué-Zapata et al., 2023, who identified persistent weaknesses in problem formulation, data analysis, and the use of statistical techniques among Peruvian students. This suggests that favorable attitudes may coexist with methodological limitations.

The most robust result of the present study corresponds to a relationship of considerable magnitude between academic self-efficacy and self-perceived research skills (r = 0.740; p < 0.001). Current research reports correlations of considerable magnitude between academic self-efficacy and variables related to strategic learning (r ≈ 0.704), indicating that students with greater confidence in their abilities often report using more effective learning strategies and stronger perceptions of their research skills (Surichaqui and Gerónimo, 2025).In nursing students, Le et al. (2025) found that research self-efficacy showed a statistically significant association with research competence (β = 0.710; p < .001), a value comparable in magnitude to that observed in the present study. However, it is important to note that research self-efficacy refers to confidence in performing research-specific tasks, whereas the present study assessed general academic self-efficacy. Likewise, research competence differs from self-perceived research skills because the latter reflects students' subjective evaluations of their abilities rather than objectively assessed performance. Therefore, the comparison should be interpreted as involving related, but not equivalent, constructs.

Complementarily, studies such as those by Chique-Aguilar et al., 2024 and Saez-Zevallos et al., 2025 emphasize that academic self-efficacy is strongly associated with the way students organize their learning behavior, maintain intrinsic motivation, and persist when facing complex tasks. Chique-Aguilar et al. show that academic self-efficacy increases the willingness to face challenges and allows for the development of skills in environments that demand autonomy and analysis, while Saez-Zevallos et al. demonstrate that academic self-efficacy is statistically associated with the use of self-regulatory strategies and academic performance.

Similarly, Cachique-Lachuma et al., 2026 showed that academic self-efficacy has been modeled as a statistical mediator in relationships among motivational and performance variables. Their findings align with the results of the present study, where academic self-efficacy emerges as the strongest predictor of self-perceived research skills.

The study by Chávez-Arizala and Soria-Quijaite, 2026 provides an interesting difference: they reported a positive correlation between academic self-efficacy and research competencies (r = 0.480; p < .001), but when other variables were introduced into an SEM model, the direct relationship was no longer significant (p = 0.157), while technological competencies and resilience maintained direct effects. These results indicate that academic self-efficacy can operate as a mediating or modulating mechanism within more complex learning systems.

Before examining the regression results, it is important to note that the correlational evidence described above provides the conceptual foundation for understanding how these variables are related within more complex analytical models. These associations suggest potential patterns through which academic self-efficacy and attitude toward research may relate to students' self-perceived research skills, thereby justifying a deeper examination of their relative contribution within the regression model.

Statistical predictors of self-perceived research skills

The results of the regression analysis of the present study show that attitude toward research is significantly associated with self-perceived research skills (β = 0.145; p < 0.001), although with a considerably smaller standardized coefficient compared to academic self-efficacy. In contrast, academic self-efficacy shows a substantially higher standardized coefficient (β = 0.678; p < 0.001), indicating its predominant statistical contribution to the model. This pattern is broadly consistent with the findings of Le et al. (2025), who identified research self-efficacy as the strongest predictor of research competence. Nevertheless, direct comparisons should be interpreted cautiously because research self-efficacy and academic self-efficacy represent different levels of specificity, and research competence is not equivalent to self-perceived research skills. Similarly, Miao et al., 2025 demonstrated that self-efficacy is statistically associated with academic performance directly (path coefficient = .332) and that, even in mediation models, the total effect remains significant. In summary, these findings support a sociocognitive interpretation of research training, in which commitment to complex tasks, persistence in the face of difficulties, and efficacy beliefs organize learning behavior.

Other applied studies show that real and concrete training experiences reinforce academic self-efficacy and, therefore, self-perceived research skills. Jones et al., 2023 demonstrated that participation in mentored research outside the classroom was positively related to research self-efficacy and outcome expectations. Similarly, Alsalim et al., 2024 reported that participation in a research symposium was associated with significantly higher levels of overall self-efficacy and self-efficacy in planning, execution, and presentation phases.

Study limitations

The cross-sectional nature of the design does not allow for establishing causal relationships among the variables. The identified associations reflect statistical covariation rather than causal influence. In non-experimental studies, the observed relationships describe patterns of association and statistical explanation, but they do not permit determining which variable originates or produces changes in another.

The generalizability of the findings is limited by the use of non-probabilistic convenience sampling, which restricts the representativeness of the sample.

The use of self-report measures may introduce social desirability biases or overestimated perceptions of one's own abilities, particularly in the case of perceived research skills, which do not constitute objective performance measures.

Another limitation relates to conceptual heterogeneity across comparative studies, as some investigations assess research self-efficacy, general academic self-efficacy, research competence, or self-perceived research skills. Although these constructs share a common component related to perceived capability, they differ conceptually and operationally. Consequently, comparisons across studies should be interpreted cautiously because similar statistical relationships may involve non-equivalent constructs.

Additionally, the high correlation between academic self-efficacy and perceived research skills (r = 0.740) may reflect common method bias, given that all variables were assessed through self-report instruments at a single time point. This pattern may indicate conceptual overlap between the instruments or a general tendency among students to evaluate themselves similarly across different domains, rather than reflecting actual differences in their competencies. Therefore, the results should be interpreted with caution, considering that the assessed skills correspond to subjective perceptions rather than objective performance measures.

An additional limitation concerns the unequal distribution of participants across academic semesters and degree programs. Students in advanced semesters may have accumulated greater exposure to research methods, statistics, scientific reading, and research-related activities. Although these variables were not included as controls because they were outside the primary objectives of the study, future research could incorporate academic semester, field of study, age, and gender as control variables to evaluate the stability of the observed associations.

Finally, although the regression model figure represents associations through arrows and standardized coefficients, this visual structure may lead to causal interpretations. In cross-sectional studies, the arrows in the model do not represent temporality or causal direction, but solely statistical explanatory relationships. Thus, the figure should be interpreted as a graphical representation of associations and statistical explanatory relationships, not as evidence of causal effects among the variables.

Implications of the study

Practical implications

The results have direct applications in university training. Academic self-efficacy showed the strongest statistical association with self-perceived research skills, suggesting that it may represent an important psychological resource in research training contexts. This suggests that instructors should design learning experiences that allow students to progressively master research methodologies, receive focused feedback, and access opportunities for academic success. Likewise, encouraging positive attitudes toward research may be associated with greater willingness to engage in scientific practices through disciplinary projects, authentic problems, and guided research experiences.

Research training should be gradual, experiential, and contextualized, incorporating practices that allow students to observe, practice, analyze, and reflect on each research procedure. It is important to note that these implications do not imply causal relationships; rather, they derive from patterns of association and statistical explanatory value, which guide instructional practices without asserting direct effects. Additionally, these implications refer to self-perceived research skills, which represent subjective evaluations rather than objective performance.

Theoretical implications

The findings contribute empirical evidence derived from correlational analyses and regression modeling, relevant for understanding the role of psychological factors in research training. The identification of academic self-efficacy as the strongest predictor of self-perceived research skills aligns with sociocognitive theory, which posits that efficacy beliefs are associated with behavior, motivation, and persistence in complex tasks. This evidence reflects patterns of association, not causality.

Likewise, the significant contribution of attitude toward research, although smaller in magnitude, reaffirms that affective-cognitive dispositions are associated with research intention and performance, consistent with the theory of planned behavior. These results suggest that explanatory models of research competence should integrate efficacy beliefs, attitudes, and formative experiences.

The theoretical implications also highlight the relevance of using observed variables and considering the multidimensional structure of the constructs when examining relationships among psychological variables. This approach allows for examining direct statistical associations without assuming latent causal mechanisms. Future research may benefit from longitudinal designs and more complex structural models that explore mediating or moderating relationships among psychological and pedagogical variables.

Institutional implications

The findings highlight the need to strengthen strategies that consolidate a university research culture. Research training pathways should be integrated from the early semesters, preventing competencies from being developed only in final courses. It is necessary to reinforce the training of research instructors in active methodologies and guided research support, given their association with students' academic self-efficacy.

Institutions should promote extracurricular research environments—such as symposiums, scientific clubs, and mentoring—and ensure access to updated scientific literature and technological resources (databases, statistical software, repositories, laboratories). Strengthening these institutional conditions may contribute to educational environments that support favorable attitudes toward research and stronger perceptions of research skills, within the limits of non-experimental evidence.

Educational implications

Based on these conceptual linkages, a key pedagogical implication of this study lies in recognizing research competencies as transversal assets within entrepreneurship education. These theoretical connections suggest the importance of designing integrated curricular strategies where analytical skills, critical thinking, and academic persistence developed through research training are not viewed solely as academic ends in themselves, but also as transferable tools for social and venture innovation environments. Fostering educational spaces that cultivate autonomous problem-solving and rigorous communication of outcomes can systematically enhance students' innovative potential without requiring the explicit or direct measurement of entrepreneurial behavior within the classroom.

The results show that students who develop more favorable attitudes toward research also tend to exhibit higher levels of academic self-efficacy. The relationship between the two variables was positive and significant (r = 0.431; p < 0.001), indicating that interest in, appreciation of, and willingness to engage in scientific activity are associated with greater confidence in one's ability to meet academic demands. This association reflects patterns of covariation rather than causal influence, consistent with the non-experimental, cross-sectional design.

Likewise, attitude toward research was significantly associated with self-perceived research skills (β = 0.145; p < 0.001). Although its explanatory contribution was moderate compared to other factors, the results indicate that positive attitudes toward research are associated with greater student engagement and participation in scientific activities. These findings refer to self-perceived research skills, which represent subjective evaluations rather than objective performance.

Academic self-efficacy showed the strongest association with self-perceived research skills (r = 0.740; p < 0.001) and was the most relevant predictor within the model (β = 0.678; p < 0.001). This suggests that students who are confident in their academic abilities are more likely to perceive themselves as competent in problem formulation, information analysis, scientific argumentation, and communication of results. These results reflect statistical associations derived from observed variables, not causal pathways.

Finally, the regression model confirmed that attitude toward research and academic self-efficacy jointly explain 56.5% of the variability in self-perceived research skills (R2 = 0.565). Consequently, the results support the importance of strengthening not only methodological knowledge but also the psychological factors associated with research training in university students. These conclusions should be interpreted within the limits of non-experimental evidence and the self-report nature of the instruments.

Recommendations

For institutional management

The following recommendations should be interpreted as practical considerations derived from the observed associations and supported by previous literature, rather than as interventions empirically validated by the present study.

Plan and implement research advising programs that connect students with experienced researchers and instructors.

Organize collaborative practice groups where students can share progress, difficulties, and strategies.

Update and strengthen scientific infrastructure—laboratories, repositories, statistical software, academic databases—to ensure accessibility for students.

Create institutional disciplinary journals where students can publish research derived from authentic projects.

Design research training programs that integrate original research experiences, strategies to strengthen academic self-efficacy, and practices that promote positive attitudes toward research.

For teaching practice

Research instructors should remain updated on methodological concepts and procedures to design original learning experiences.

Incorporate real-world projects—case studies, micro-research projects—that allow students to apply the scientific method and disseminate findings in institutional venues.

Strengthen formative feedback by offering specific, timely, and process-oriented comments that enhance students' self-perceived research skills and encourage metacognitive reflection through research notes, error analysis, and self-regulation strategies.

For future research

Conduct longitudinal studies to observe how academic self-efficacy, attitude toward research, and self-perceived research skills evolve throughout university education.

Explore additional psychological and contextual variables such as academic resilience, intrinsic motivation, digital skills, and university climate.

Evaluate the impact of different instructional methodologies on research training.

Analyze mediating or moderating relationships among psychological and pedagogical variables using structural equation modeling.

Future studies could also simultaneously examine academic self-efficacy, research self-efficacy, entrepreneurial self-efficacy, research competencies, and entrepreneurial competencies in order to better understand their interrelationships. Longitudinal and intervention-based designs would be particularly valuable for examining how these constructs evolve over time and contribute to innovation-oriented educational outcomes.

Future research should evaluate the empirical discriminant validity between academic self-efficacy and self-perceived research skills using variance-based structural equation modeling (PLS-SEM) to assess the Heterotrait-Monotrait ratio (HTMT) or confirmatory factor analysis (CFA). Additionally, to mitigate potential common-method bias from self-reported data, future designs should incorporate objective behavioral metrics or rubric-based performance assessments to verify whether self-efficacy beliefs align with demonstrated research competence.

Furthermore, future studies should utilize hierarchical regression and multigroup structural models to account for academic progression. By controlling for academic semester or study level, researchers can isolate the incremental variance of specific research competencies beyond general academic self-efficacy. Balanced longitudinal or cross-sectional designs across academic stages would further clarify whether the influence of research skills evolves as students advance through higher education.