Academic Integration as the Main Driver of Student Retention: A Multidimensional Analysis of Educational Ecosystems in Private Universities in Northern Peru

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

After checking and reading your manuscript (societies-4055650) I think you should perform major revisions, namely:

-The sampling (convenience-based, most probable) from one university severely limits generalizability. The high concentration in Industrial Engineering (60%) and Nursing (20%) may skew results and is not representative of broader populations. Therefore, claims about private universities in northern Peru are not supported.

-The cross-sectional design (only single point in time measures) prevents causal claims.

-The dependent variable is self-reported more as intention to persist rather than objective retention data. The well-documented intention-behavior gap means students with high integration scores might still drop out due to external factors like financial shocks.

-All data (both independent and dependent variables) come from the same self-report questionnaire at the same time. This can artificially inflate relationships between variables due to respondent mood, social desirability bias, or consistency motifs.

-While appropriate for exploratory research, the paper lacks advanced techniques like Structural Equation Modeling (SEM) to test mediation/moderation effects, despite the theoretical framework suggesting them. The paper proposes academic integration as a "proximal mediator" but doesn't actually test mediation.

-There are no static limits of VIF (such as 5 or 10). You should reconsider (dynamic thresholds) depending on the model's R^2 value (https://doi.org/10.4172/2161-1165.1000227 and https://books.google.ro/books?hl=ro&lr=&id=Us4YE8lJVYMC). 

-There are signs of weak operationalization of institutional support. This construct had the lowest mean score (3.70) and borderline reliability (α=0.75). Items measure student awareness/utilization rather than actual quality or availability of support. The ambiguity makes it unclear whether low scores reflect poor institutional offerings or poor student awareness.

-There is so much overgeneralization in Discussion and Implications. Given the sampling limitations, the manuscript's language about implications for private universities in northern Peru is too broad. Some implications feel generic (e.g., invest in well-being programs) and could be more tailored to Peru's specific context (poverty, regional challenges). More attention is needed and more restraint in claims about generalizability.

-Not enough analysis of non-significant findings - some predictors (e.g., self-regulation skills, clarity of expectations, academic preparation) were non-significant, but the discussion doesn't deeply explore why. This represents a missed opportunity for theoretical insights about what types of integration matter most. Therefore,
a superficial treatment of unexpected null findings limits theoretical contribution.

-Limited theoretical novelty - meaning that While the literature review is comprehensive, the core ideas aren't entirely novel. The paper heavily relies on established frameworks (Tinto, SDT) without introducing new theoretical constructs. Hypotheses are predictable extensions of prior work, and findings largely confirm existing meta-analyses.

-There is also a notable lack of subgroup analysis. The sample's heterogeneity (38.3% first-generation students, mostly young adults, 99.1% single) isn't analyzed through subgroup comparisons. No exploration of whether retention factors differ by student characteristics, field of study, or first-generation status. Or this means missing potential insights into differential effects across student populations.

 

I wish you all the best!

Comments on the Quality of English Language

While the English is largely correct, there are a few minor stylistic points that could be polished during the final editing stage to achieve native-level fluency and conciseness. The authors should use Grammarly or a similar tool and aim for the maximum score possible.

Author Response

 

a) Comments raised (changes implemented in the manuscript)

Generalisability/sampling from a single university + bias by degree programme (60% industrial engineering; 20% nursing)

Action: The scope of the manuscript was corrected to be consistent with a sample from a single institution.

Where:

Title: now indicates "a private university" (not "private universities").

Abstract: specifies "a private university in northern Peru" and softens the closing statement ("similar regional contexts").

Objective (Introduction) and Methods (design and sample): all statements of "representativeness" are removed and the lack of statistical generalisation is stated.

Results (sample profile): an explicit note has been added stating that disciplinary concentration may affect the pattern of associations and limits extrapolation.

Discussion/Implications/Conclusions: the language has been adjusted to avoid generalisations about "all private universities in northern Peru".

Cross-sectional design: avoids causal inferences

Action: The language of association/prediction (not causation) was reinforced and consolidated as a central limitation.

Where: Limitations, Discussion, and Conclusions (explicitly repeating "predictive associations," "cross-sectional, intention-based").

Dependent variable is intention to persist (not objective retention) + intention–behaviour gap

Action: The limitation was expanded and the intention–behaviour gap and the possibility of external shocks (financial/family/health) that break the intention→behaviour translation were explicitly incorporated.

Where: Limitations and recommendations for future research (link to administrative records).

Common method bias (same survey, same time, self-report)

Action: An explicit limitation of common-method variance was added, listing plausible mechanisms (mood, social desirability, consistency) and proposing future remedies (multi-source, temporal separation, marker variables).

Where: Limitations.

VIF: avoid rigid cutoffs (5/10); interpret contextually (dependent on auxiliary R²)

Action: The analysis section was rewritten to state that there is no universal threshold, that VIF depends on auxiliary R², and that it is reported transparently along with stability diagnostics.

Where: Statistical analysis and results paragraph where VIF is reported (language such as "meets fixed threshold" is avoided).

Operationalisation of "Institutional support": risk of measuring awareness/use rather than actual quality

Action: The construct was redefined as perceived visibility/awareness & utilisation of services + institutional care (not "objective quality"), and an interpretation was added that low scores may reflect low visibility/use in addition to availability/quality.

Where:

Definitions (Literature Review),

Study variables (Methods),

Interpretive text in Results,

Appendix A (Table A1) updated so that the conceptual domain and scoring reflect "support/visibility".

Overgeneralisation in discussion and implications (too broad for "private universities in northern Peru")

Action: The language was "revised": now it refers to "sampled institution" and "comparable regional contexts", and generalised prescriptions are avoided.

In addition: It became more contextual by using the socioeconomic profile described in the sample (without inventing data) so that the implications are less generic.

Where: Practical implications and Conclusions.

Little discussion of non-significant findings.

Action: The interpretation of the single non-effect of personal factors in the joint model was expanded: it is explained as overlapping with academic integration (theoretical/empirical proximity), and future research is proposed to disaggregate sub-dimensions and interaction by subgroups.

Where: Discussion (main findings) and Theoretical implications.

b) Refuted or justified observations (not completely removed; explained and/or treated as a limitation)

"Sampling from a single university invalidates claims about 'private universities in northern Peru'."

Justification: Sampling cannot be "corrected" a posteriori without new data.

What was done instead: The scope (title + claims) was corrected and the limitations/external validity section was strengthened.

Editorial outcome: The criticism is addressed by precision of inference, not by changing the design.

"Lack of subgroup analysis (first-generation, field of study, etc.)"

Justification: Without a database (and without an original analytical plan), it is not possible to perform subgroup comparisons or interaction models without the risk of inventing results.

What was done: It was incorporated as an explicit limitation and as a line of future research (multi-group/SEM, adequate power, multi-institutional sample).

“Limited theoretical novelty (dependence on Tinto/SDT; predictable hypotheses)”

Justification (partial refutation): The manuscript does not attempt to introduce new constructs; its contribution is integrative and contextual: it quantifies reliable composite constructs and estimates unique contributions in a regional context outside the capital, connecting it with SDG 4.

What was done: The contribution was explicitly reframed to avoid promises of strong theoretical novelty and position it as an empirical/integrative contribution.

“Apply dynamic VIF thresholds dependent on R²”

Justification: Without redoing the model or introducing new ad hoc rules, the most rigorous approach is not to use fixed cutoffs and to report VIF + tolerance + stability, which is what was implemented.

What was done: The "fixed threshold" logic was eliminated and it was made clear that VIF is interpreted contextually, reporting values and avoiding undue inferences.

Reviewer 2 Report

Comments and Suggestions for Authors

Dear authors,

The paper addresses a relevant topic, student retention representing a major concern for higher education institutions, especially in the context of public or private universities and educational ecosystems in emerging regions, such as northern Peru. The proposed multidimensional analysis, with a focus on academic integration, offers an interesting perspective and is well-anchored in the specialized literature. The study is timely and potentially valuable from both a theoretical and practical point of view. However, in order to increase the clarity, methodological rigor and scientific contribution of the paper, a series of revisions and clarifications are necessary, detailed in the observations below, which can significantly contribute to the improvement of the final version of the article.

1. In the abstract, a careful review of typos is necessary, as even in the abstract there are words written with an incorrect hyphen. Also, SDG 4 is used only as an abbreviation, so it is recommended that the full name be presented at the first mention, followed by the abbreviation. It is also not necessary to include the values ​​of the regression equations in parentheses in the abstract (they can be presented exclusively in the results section).
2. At the end of the introduction, the paragraph describing the structure of the paper is missing, which would help the reader understand the organization of the article.
3. It would be useful if section 1.1 were transformed into a separate section, such as “2. Literature Review”. A short transition paragraph should be inserted between section 2 and subsection 2.1, specifying what aspects will be discussed in the following subsections.
4. The subsections in the Literature Review are too succinct. Beyond the presentation of theories, when discussing relationships or factors that influence student retention, several empirical studies should be included, with: methodologies used, main results, possible limitations. Based on the synthesis of these studies, the gap in the literature should be more clearly highlighted, as well as the hypotheses or research questions to be tested.
5. In section 2.3 it remains unclear what types of variables are used, how they are operationalized, by which items they are measured, what type of scale was used and what the basic questionnaire is. This information should be clarified, and the research instrument included as an annex.
6. In section 2.5 it is mentioned that "Data collection was carried out during the 2024 semester", but it is not clear which semester this is about or what the exact period of the year was. Given that the data were collected over a relatively long period and via an online questionnaire, the possibility that the timing of completion (beginning of the semester vs. end of the semester/approaching exams) may have influenced respondents’ attitudes should be discussed.
7. After the results section, a brief transition description would be useful before moving on to the following subsections to guide the reader.
8. In the case of Tables 1–5, it is not sufficiently explained how and why the means were calculated. For example, in Table 2, the item Clarity of academic/professional goals has a mean of 4.2, but the interpretation of this value and its relevance are not discussed, especially considering that the items are measured on a Likert-type agree/disagree scale.
9. In section 3.4, where the proposed models are tested, the tables with the estimated regression models, including the relevant coefficients and indicators, should be explicitly presented.
10. Greater care is needed when formulating conclusions. R² (or adjusted R²) indicates the proportion of variation explained by the entire model, not the contribution of an individual variable. Therefore, statements such as “academic integration emerges as the strongest predictor” based on the adjusted R² value are methodologically incorrect. The same observation applies to other similar interpretations, which need to be reviewed and corrected.

Comments on the Quality of English Language

The overall quality of the English language is acceptable; however, the manuscript would benefit from a careful language revision. There are several issues related to grammar, word choice, punctuation, and typographical errors, including the inconsistent use of hyphenation and occasional awkward phrasing. Some sentences would benefit from restructuring to improve clarity and flow. A thorough proofreading or professional language editing is recommended prior to publication.

Author Response

 

Comment 1

In the abstract, a careful review of typos is necessary… SDG 4 should be written in full at first mention… and regression equation values should not appear in the abstract.

Response:
We revised the Abstract to (a) correct typographical issues and improve hyphenation/wording, (b) introduce SDG 4 by its full name at first mention as “Sustainable Development Goal 4 (SDG 4)”, and (c) remove the regression equation/coefficient values that were previously shown in parentheses. The Abstract now reports the key substantive takeaways and keeps detailed regression estimates exclusively in the Results (Table 6), as recommended.

Implemented in: Abstract; Results/Table 6.

Comment 2

At the end of the introduction, the paragraph describing the structure of the paper is missing.

Response:
We added a short paper-organization paragraph at the end of the Introduction. This paragraph explicitly guides the reader through the structure of the manuscript (Literature Review → Methods → Results → Discussion → Conclusions), improving navigability and coherence.

Implemented in: End of Section 1 (Introduction).

Comment 3

Transform section 1.1 into a separate section (e.g., “2. Literature Review”) and insert a transition paragraph before subsection 2.1.

Response:
We restructured the manuscript so that the former Section 1.1 is now a standalone Section 2: Literature Review. Consequently, subsequent sections were renumbered to preserve logical flow (Methods → Results → Discussion → Conclusions). In addition, we inserted a transition paragraph immediately after the Section 2 heading to explain what the upcoming subsections cover and how the literature review leads to the study gap and hypotheses.

Implemented in: Section 2 (Literature Review), including a new transition paragraph before Section 2.1; global renumbering of sections.

Comment 4

The Literature Review subsections are too succinct. Include empirical studies (methodologies, results, limitations), clarify the gap more clearly, and strengthen hypotheses/research questions.

Response:
We expanded the Literature Review substantially. Beyond presenting theories, we now synthesize empirical evidence for each major domain (personal factors, academic integration, institutional support), explicitly summarizing:

• typical methodological approaches used in the literature (e.g., cross-sectional survey, meta-analytic evidence, intervention reviews),

• main findings in qualitative terms (without inventing new quantitative results), and

• common limitations (e.g., reliance on self-reports, intention-based outcomes, context dependence).

Based on this synthesis, we strengthened the articulation of the research gap (joint estimation of the constructs in a regional private higher education setting) and refined the presentation of the hypotheses so they are more clearly derived from the reviewed evidence and aligned with the analytical strategy.

Implemented in: Sections 2.2–2.5 (expanded literature + clearer gap + hypotheses).

Comment 5

In section 2.3 it is unclear what variables are used, how they are operationalized, which items measure them, what scale is used, and what the questionnaire is. Include the instrument as an annex.

Response:
We clarified the study variables and their operationalization in the Methods. Specifically, we now state:

• which variables are treated as composite constructs,

• which items/indicators belong to each construct,

• the response scale (five-point Likert agreement scale), and

• the scoring approach used to compute composites (item averaging).

In addition, to address the request for the “basic questionnaire,” we included the core measurement instrument as an annex: Appendix A, containing (i) Table A1 mapping constructs to indicators and scoring, and (ii) a concise list of the core scale items used to compute each composite.

Implemented in: Section 3.3 (Study variables) and Appendix A (Table A1 + core item list).

Comment 6

“Data collection was carried out during the 2024 semester” is unclear. Specify which semester/period; discuss the possibility that timing (early vs late semester) influenced attitudes.

Response:
We revised the Data collection procedure to require an explicit statement of the data collection window, specifying both the semester and exact months. Because the original text did not report the precise window, we inserted a clear slot to report it as “Semester [I/II], [Month–Month] 2024” so the authors can populate the exact period accurately.

Additionally, we added a dedicated methodological note acknowledging that, because the online survey remained open over multiple weeks, response timing (early-semester adjustment vs end-of-semester/exam pressure) could have influenced attitudes. We explicitly state this as a potential limitation and recommend that future work record completion timing or conduct sensitivity analyses.

Implemented in: Section 3.5 (Data collection procedure) (date-window clarification + timing/measurement variability discussion).

Comment 7

After the results section, add a brief transition description before moving to the following subsections.

Response:
We added guiding transitions to improve readability:

• At the beginning of Results, we inserted a short roadmap describing the order of reporting (sample → measurement model → descriptives → regression).

• At the end of Results, we inserted a short transition indicating that the next section interprets findings within the theoretical framework and discusses implications (including SDG 4) while acknowledging analytical limits.

Implemented in: Start of Section 4 (Results) and immediately before Section 5 (Discussion).

Comment 8

Tables 1–5: insufficient explanation of how/why means were calculated; interpret means on a Likert agree/disagree scale (e.g., mean 4.2).

Response:
We strengthened the explanation of descriptive statistics in two ways. First, in Statistical analysis, we explicitly define how item means were computed (arithmetic average of 1–5 responses) and how to interpret them on an agreement scale (values above 3 indicate a tendency toward agreement; values approaching 4–5 indicate stronger endorsement). We also clarify what % Agree (4–5) represents.

Second, we added brief interpretive statements in the Results narrative around Tables 2–5 to contextualize what the reported means imply substantively (without repeating the entire table or over-reporting statistics).

Implemented in: Section 3.6 (Statistical analysis) and interpretive text accompanying Tables 2–5 in Section 4.3.

Comment 9

In section 3.4, where the proposed models are tested, tables with estimated regression models (coefficients and indicators) should be explicitly presented.

Response:
We ensured that the regression model results are presented explicitly and transparently. The hypothesis-testing subsection now clearly points the reader to the regression output table (Table 6), which reports the relevant coefficients and indicators (B, SE, β, t, p), as well as diagnostic/quality indicators (VIF) and each predictor’s unique contribution (sr²/ΔR²) within the same model specification.

Implemented in: Section 4.4 (Hypothesis testing: multiple regression model) and Table 6.

Comment 10

Conclusions must be revised: R²/adjusted R² reflects the whole model, not an individual variable. Statements like “strongest predictor” based on adjusted R² are incorrect.

Response:
We fully agree and revised the manuscript to correct this methodological interpretation. We removed any wording that implied that adjusted R² indicates the contribution of an individual predictor. Where comparative language is needed (e.g., “largest/strongest association”), it is now grounded in appropriate within-model indicators such as the standardized coefficient (β) and the unique explained variance (sr²/ΔR²) reported for each predictor in Table 6. We also added a short comparative summary section that explicitly states that comparisons are derived within the same model, not across separate R² values.

Implemented in: Section 4.5 (Comparative summary of the regression model), and reinforced in Section 6 (Conclusions).

Closing note

Once again, we appreciate the Reviewer’s feedback. We believe these revisions substantially improve the manuscript’s clarity, structure, and methodological rigor, and they enhance the interpretability of results and their implications for SDG 4 within the empirical limits of the study.

 

Reviewer 3 Report

Comments and Suggestions for Authors

Theoretical background: The literature review is comprehensive and well-situated.

Research design: A critical methodological flaw exists. The study validates three higher-order constructs (Personal Factors, Academic Integration, Institutional Support) via psychometric analysis but does not use these composite constructs in the regression. Instead, it decomposes them into individual items. This disconnect between measurement validation and analytical model undermines the study's theoretical and methodological coherence. The research design is therefore not aligned with its multidimensional analytical goal.

Discussion: Built upon an analytical foundation conceptually misaligned. The arguments about the "relative contribution" of each theoretical dimension are based on models that do not test these dimensions as unified constructs, making comparisons of adjusted R² values across separate models problematic and the conclusion about a "hierarchy" of predictors potentially misleading.

Conclusions: The conclusions regarding the hierarchical importance of the three dimensions are not supported by the results. The validated composite scores for each construct should be entered simultaneously into a single regression model. The current analysis provides insights only into which specific items within each domain are significant.

 

So, there is misalignment between Construct Validation and Analysis
The manuscript performs a psychometric validation (Table 1) to create reliable composite scores for three theoretical constructs. However, the regression analysis (Section 3.4) abandons these validated composites and it runs three separate regressions, each using the individual items.

This approach :

- Does not test the multidimensional model. The title and abstract promise an analysis comparing constructs, but the presented models test them in isolation.

- Invalidates the comparative interpretation of R². The R² values come from different models with different predictors. The only way to fairly compare the unique variance explained by each construct is to enter their composite scores into a single regression.

- Raises the question of why the constructs were validated.

Required Major Revision:
- Create Composite Variables: Calculate mean or sum scores for each of the three validated constructs (Personal Factors, Academic Integration, Institutional Support) for each participant.

- Perform a Single Multiple Regression with these three composite scores as independent variables.

- Revise Results and Discussion.

Author Response

We agree with the reviewer that the initial version created validated constructs (Table 1) but then estimated separate regressions using individual items, which limits construct-level inference and makes model fit comparisons inappropriate. To address this, we computed composite scores (item means) for each validated construct—personal factors (PF_comp), academic integration (AI_comp), and institutional support (IS_comp)—as well as the outcome intention to persist (IP_comp). We then estimated a single multiple linear regression model with the three composites entered simultaneously (Table 6). This specification evaluates the multidimensional framework in one model and allows fair comparison of predictors using standardized coefficients and unique contributions (sr²/ΔR²) within the same model, rather than comparing R² across different model specifications. The revised results show academic integration as the strongest predictor, followed by institutional support, while personal factors do not retain a unique effect once shared variance is controlled. This revision directly resolves the alignment issue between construct validation and the main inferential analysis.

Reviewer 4 Report

Comments and Suggestions for Authors

Greetings!

Thank you for the opportunity to review your manuscript.

Generally, your manuscript addresses an important and timely issue in higher education research by examining student retention through a multidimensional framework that integrates personal factors, academic integration, and institutional support within the context of private universities in northern Peru. The study is theoretically grounded, methodologically rigorous, and supported by a strong body of recent and relevant literature. The empirical results are well presented and offer meaningful insights, particularly regarding the central role of academic integration in shaping students’ intention to persist, making a valuable contribution to the literature on student retention in Latin American higher education contexts.

At the same time, several areas could benefit from revision to strengthen clarity, coherence, and alignment between the conceptual framing and the analytic approach. Note, the recommendations are a refinement issue, not a conceptual weakness of the study. Therefore, first, the Introduction and Literature Review are comprehensive but, at times, overly dense, with some repetition across the theoretical frameworks. Similar incidents of repetition were also noted in the discussion area (See note below). Streamlining these sections by more clearly articulating the distinct role each framework plays in the proposed model and reducing overlapping explanations would improve readability and sharpen the manuscript’s theoretical contribution. Clarifying academic integration as the central organizing construct earlier in the paper may also help guide readers more effectively through the analysis.

For example:

• Reduce redundancy across theory sections.
• More clearly position academic integration as the central organizing construct, rather than one of several equally weighted frameworks.
• Streamline citations where multiple sources support the same point.

Second, while the use of three separate regression models is justified and clearly executed, the manuscript would benefit from a more explicit methodological rationale for this choice in relation to the stated multidimensional model. As currently written, the framing suggests potential interactions or mediating relationships among dimensions that are not empirically tested. Strengthening the justification for independent models and more clearly acknowledging the analytical boundaries this choice imposes would enhance methodological transparency. Relatedly, the discussion could more consistently distinguish between predictive associations and causal interpretations, particularly given the cross-sectional design and reliance on intention-based measures.

Lastly, the Discussion and Conclusions sections could be tightened further. Several findings are reiterated across sections with limited additional interpretation. Reducing redundancy and focusing more explicitly on the implications of the hierarchical pattern observed (academic integration, followed by personal factors, then institutional support) would strengthen the manuscript’s analytical depth. The policy and SDG 4 implications are relevant and appropriate, but would be more compelling if more clearly tied to the study's empirical scope and limitations. Addressing these points through revision would significantly enhance the clarity, impact, and overall quality of an already strong manuscript.

For example, as it relates to the discussion area:

• Reduce repetition of statistical results already presented
• Focus discussion on interpretive insights, especially why academic integration outweighs institutional support in this context
• Tighten SDG 4 alignment, so it is analytical rather than descriptive. This is very a very important element!

Similarly, with the conclusions:

• Narrow claims about policy implementation
• Re-emphasize limitations when drawing implications
• Avoid implying causal pathways not directly tested

On a related note, the English is generally grammatically correct, but the manuscript could benefit from a revisit for clarity and readability:

• Shorter sentences
• Reduced nominalization
• Less repetition of key phrases (for example, “sustainability of educational ecosystems”).

 

Therefore, I recommend that the manuscript be accepted after the major revisions are made.

Best wishes!

Comments on the Quality of English Language

Greetings!

The manuscript is written in clear and generally correct academic English, and the authors demonstrate a strong command of disciplinary terminology and scholarly style. However, revisions would improve clarity and readability, particularly by reducing sentence length, limiting repetition of key phrases, and tightening prose in sections where multiple ideas are introduced simultaneously. In several places, long and complex sentences obscure otherwise strong arguments. Minor stylistic editing focused on concision and flow would help ensure that the manuscript’s contributions are communicated as effectively as possible to an international readership. Please note that the suggested revisions are editorial in nature and do not detract from the overall strength and quality of the work.

Author Response

In response to your comments, the manuscript was thoroughly revised and a refined version is attached: (i) The Introduction and Literature Review were simplified by eliminating redundancies and theoretical overlaps, and it was made clear from the outset that academic integration is the central organising construct, differentiating the specific role of each framework within the model (academic integration as the core, and personal factors/institutional support as complementary dimensions); in addition, citations were streamlined when several sources supported the same argument, retaining only the key references per idea. (ii) In Methodology, a more explicit justification was incorporated for the choice of independent regression models in relation to the multidimensional approach of the conceptual framework, and the analytical scope of this decision was clearly recognised (e.g., that mediations/interactions are not tested) along with the limitations derived from the cross-sectional design and intention-based measures, reinforcing the distinction between association/prediction and causality. (iii) In Discussion and Conclusions, the repetition of already reported results was reduced and the interpretation of the observed hierarchical pattern (academic integration > personal factors > institutional support) was reinforced, with particular emphasis on why academic integration might dominate in this context without introducing untested causal inferences. Likewise, alignment with SDG 4 was strengthened analytically (linking findings, empirical scope, and limitations), and policy statements were moderated to be consistent with the study design. Finally, English style editing was performed to improve readability (shorter sentences, less nominalisation, and reduction of key phrase repetitions).

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

When considering v1, the initial review letter, and v2, this second version is clearly improved, but not consistently.
Your responses and manuscript updates do not fully satisfy the underlying concern (e.g., robustness, justification, generalizability, limits).
The methods are clearer than in v1, but some assumptions remain implicit, some choices are justified narratively rather than analytically, and limitations are acknowledged, but not always fully operationalized.

I wish you all the best!

Sincerely,
One of the reviewers

Comments on the Quality of English Language

While the English is largely correct, there are a few minor stylistic points that could be polished during the final editing stage to achieve native-level fluency and conciseness. The authors should use Grammarly or a similar tool and aim for the maximum score possible.
Additionally, some rebuttal-style phrasing still feels more suitable for a response letter than for the manuscript.

Author Response

Dear Reviewer,

Thank you for your careful reading of our revised manuscript and for acknowledging the improvements from v1 to v2. We also appreciate your clear articulation of the remaining underlying concerns (robustness, analytical justification, generalizability, and the operationalization of limitations). We took your feedback very seriously and made a concerted, comprehensive effort to address these issues consistently throughout the manuscript.

In this final revision, we focused on strengthening coherence and methodological transparency without adding new citations, constructs, or analyses. Specifically, we implemented the following changes across the Abstract, Introduction, Methods, Discussion, and Conclusions:

1. Making implicit assumptions explicit (interpretation of coefficients)
   We clarified that the regression coefficients are interpreted as conditional (unique) associations under a joint specification, rather than as bivariate relationships or causal effects. We also explained that a non-significant predictor in the joint model reflects limited unique variance beyond the other dimensions, not necessarily the absence of a relationship in isolation.

2. Shifting from narrative to analytical justification (model specification and rationale)
   We explicitly stated the model specification and the analytical rationale for entering predictors simultaneously, emphasizing that this choice allows us to separate unique from shared variance among conceptually related dimensions. This directly addresses the concern that some choices were previously justified narratively rather than analytically.

3. Strengthening robustness using already-reported diagnostics and effect-size metrics
   Without introducing new analyses, we reinforced robustness by consistently grounding interpretation in the already reported metrics and diagnostics (e.g., standardized coefficients, sr²/ΔR², multicollinearity indices, and Durbin–Watson), and by explicitly acknowledging construct overlap (e.g., the strong correlation between personal factors and academic integration) to avoid over-interpretation of single statistics.

4. Tightening claims and aligning generalizability across sections
   We revised wording to ensure that claims remain within the empirical scope of a cross-sectional, self-report, single-institution study in northern Peru using an intention outcome. We aligned the Abstract, Discussion, and Conclusions to emphasize context-bound, predictive associations rather than broader statistical generalization or causal inference.

5. Operationalizing limitations into concrete research requirements
   We revised the limitations section to specify what data and designs would be needed to address key constraints (e.g., longitudinal follow-up, multi-institution replication, linking survey measures to administrative retention outcomes, and testing indirect/conditional pathways only once temporal ordering and behavioral outcomes are available).

We hope these revisions fully address the underlying concerns you raised and improve the manuscript’s consistency, analytical clarity, and interpretive rigor. Thank you again for your time and for pushing us to strengthen the work to a higher academic standard.

Sincerely,
The Authors

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have taken the provided recommendations into account and have made the necessary revisions. I am satisfied with the changes, and the quality of the work has improved considerably.
However, the authors  should verify that the ‘Conclusions’ section does not appear twice. I am not sure whether this duplication is caused by the track changes view on my side or if it is an actual error.

Comments on the Quality of English Language

The overall quality of the English language is acceptable; however, the manuscript would benefit from a careful language revision. There are several issues related to grammar, word choice, punctuation, and typographical errors, including the inconsistent use of hyphenation and occasional awkward phrasing. Some sentences would benefit from restructuring to improve clarity and flow. A thorough proofreading or professional language editing is recommended prior to publication.

Author Response

In fact, it is due to exchange controls. The same applies to section 4, which appears to be duplicated, but is not; it is due to exchange controls.

Reviewer 3 Report

Comments and Suggestions for Authors

Major issues were addressed.

Author Response

Thank you for your valuable contributions!

Reviewer 4 Report

Comments and Suggestions for Authors

Greetings!

In a previous review of the manuscript, there were concerns related to sentence length, repetition, and overall clarity for an international readership. In reviewing this version, I find that these concerns have been largely and effectively addressed. The manuscript is now clearer in academic English, with improved sentence structure, reduced redundancy, and better control of conceptual density across sections. The arguments, then, are communicated more directly, and the flow of the writing has been strengthened without altering the substance of the analysis. The literature review, methods, results, and discussion are all well-organized. I did note minor editorial refinements, such as small opportunities to further tighten some areas, and a minor typo in the Conclusions heading.

From my perspective, there is a slight conceptual repetition in the discussion of self-determination meta-analyses and motivational pathways. As a suggestion, trim repeated references to common methodological limitations (cross-sectional designs, intention-based outcomes). Again, the repetition of the word, Conclusions at the end…Was that just a correction, and will be addressed by the editorial team? These are editorial refinements and do not detract from the strength of the revisions to the manuscript. Well done!

Best wishes!

Comments on the Quality of English Language

Greetings!

The manuscript is written in clear and generally correct academic English, and the authors demonstrate a strong command of disciplinary terminology and scholarly style. However, revisions would improve clarity and readability, particularly by reducing sentence length, limiting repetition of key phrases, and tightening prose in sections where multiple ideas are introduced simultaneously. In several places, long and complex sentences obscure otherwise strong arguments. Minor stylistic editing focused on concision and flow would help ensure that the manuscript’s contributions are communicated as effectively as possible to an international readership. Please note that the suggested revisions are editorial in nature and do not detract from the overall strength and quality of the work.

Author Response

Corrected paragraph:

“Several limitations must be considered when interpreting the results. First, two methodological features limit causal inference: the cross-sectional design and reliance on self-reported intentions rather than observed behaviors. While regression models identify predictive associations, they cannot establish temporal precedence or rule out reverse causality. Moreover, although intentions are important predictors of behavior, gaps between intention and action are well-documented. Future studies should employ longitudinal designs that track students over several semesters and link survey data with administrative enrollment records to examine how changes in motivation, integration, and support relate to actual persistence decisions”


Thank you for the recommendations that helped enrich the research.

 

Round 3

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

You have significantly improved this manuscript.

The reduction from 36 to 29 references is an obvious quantitative downgrade. Still, let's forget about quantity and focus on quality.

I wish you all the best!

Comments on the Quality of English Language

While the English is largely correct, there could be a few minor stylistic points that could be polished during the final editing stage to achieve native-level fluency and conciseness.