Online Education Management: A Multivariate Analysis of Students’ Perspectives and Challenges during Online Classes

Round 1

Reviewer 1 Report

The document "Online education management: a multivariate analysis of students' perspective and challenges during online classes" describes how to improve online education management.

 

This article defines and explains through the research methodology using structural equation models of partial least squares, how the specific communication problems of online education; the skills of teachers to teach online classes; the quality of online education; the stress students feel during online education; and the technical requirements for online education influence the students of online education.

 

The objective of the study is indicated on several occasions, the research question is not clear, so it must be improved.

 

Although, as the authors indicate, there is enough literature focused on the perspective of students, different studies have studied the most challenging factors proposed by the authors in higher education, so the hypotheses are not supported by a theoretical framework.

 

More information should be given on the questions in the questionnaire, as it is not clear that the proposed questionnaire is supported by a review of the literature, so it needs to be improved.

 

The authors present the model proposed in figure 1 and figure 2, but it is necessary to indicate which are the hypotheses to facilitate the understanding of the text of the article.

 

The Number of cases must be justified; Example: 200 questionnaires, are they enough? why?

 

More information about the model should be given; Example: HTMT and Structural model result, It should also be indicated when the results are considered correct (justifying with references)

 

After reading the discussion and conclusions section, it is not clear what the conclusions and recommendations arising from the work. Therefore, the conclusions must be explained more clearly, and the recommendations and strategies to be carried out by the high education managers should be indicated.

 

Likewise, due to the importance of the work, it is necessary to point out more clearly the limitations of the study, for which reason a subsection of conclusions should be developed.

 

Author Response

Dear Reviewer,

Thank you for your observations and for the opportunity to improve our manuscript!

We are very grateful for taking the time to analyze the paper and make very useful, encouraging and thoughtful comments and recommendations.

We have read the evaluation and, based on the review reports, we performed revisions of our manuscript, as requested, highlighted with red into the manuscript.

 

1. The objective of the study is indicated on several occasions, the research question is not clear, so it must be improved.

 

Response: Our research is focused on hypotheses, not on research questions. But we understand that adding a general research question that comprises the nine hypotheses we addressed in the paper offers a clearer perspective for the readers. Thus, we added a paragraph at the beginning of Heading 3: The most important research question we wanted to answer refers to the measures that higher education managers might take in order to ensure a high-quality education for students. With this question at the core of our research, we conducted our analysis starting from the following nine hypotheses.

 

2. More information should be given on the questions in the questionnaire, as it is not clear that the proposed questionnaire is supported by a review of the literature, so it needs to be improved.

 

Response: Thank you for the recommendation. Indeed, table 1 that describes the questionnaire missed a last column. Thus, we added the column Source for explaining the source of our questions. Each source is explained in detail in Section 2. Literature review for each of the main topics in the questionnaire.

 

3. The authors present the model proposed in figure 1 and figure 2, but it is necessary to indicate which are the hypotheses to facilitate the understanding of the text of the article.

Response: At the beginning of Section 3. Research Methodology and Hypotheses development we presented all nine hypotheses. Because figure 1 and 2 (now 2 and 3 because another reviewer asked us to add a new figure before them) are explaining the correlations of the hypothesis, we followed your recommendation and added an explanation before the figures: Figure 2 illustrates the model we proposed for the present research (H1: COM -> QLT; H2: COM -> STRS; H3: PSK -> COM; H4: PSK -> QLT; H5: PSK -> STRS; H6: STRS -> QLT; H7: TECH -> COM; H8: TECH -> QLT; H9: TECH -> STRS) and Table 1 details the model’s constructs, items and codes.

 

4. The Number of cases must be justified; Example: 200 questionnaires, are they enough? why?

Response: We added the reference Kock, N.; Hadaya, P. Minimum sample size estimation in PLS‐SEM: The inverse square root and gamma‐exponential methods. Inf. Syst. J. 2018, 28, 227–261. https://doi.org/10.1111/isj.12131.

Thus, the minimum sample is, in accordance with the 10-times rule - 40 respondents (10 multiplied by the number of relations that influence QLT which is 4).

We added the text: After eliminating the incomplete surveys, we remained with 200 valid questionnaires, which is in accordance with the minimum sample required by this method [37].

 

5. More information about the model should be given; Example: HTMT and Structural model result, It should also be indicated when the results are considered correct (justifying with references)

Response: Thank you for the recommendation. We added, as the other reviewers also requested it, the calculation for HTMT confidence intervals bias corrected. Thus, we added the Table no 6 and the explanations. Also, we added two more references for explaining the results:

Ringle, M. HTMT discriminant validity. Forum SmartPLS. Available at https://forum.smartpls.com/viewtopic.php?t=3616 (accessed on 09 January 2023).

Henseler, J.; Ringle, C.M.; Sarstedt, M. A new criterion for assessing discriminant validity in variance-based structural equation modeling. J. of the Acad. Mark. Sci. 2015, 43, pp. 115–135. https://doi.org/10.1007/s11747-014-0403-8

For checking the discriminant validity of the model, we calculated Heterotrait-monotrait ratio (HTMT) confidence intervals bias corrected in table 6, as suggested by Ringle [43]. The confidence intervals in table 6 show a good discriminant validity between the constructs even if the correlation between STRS and COM has a borderline value for HTMT (0.909). As Ringle [43] and Henseler et al. [44] state, if the confidence intervals for HTMT do not include the value 1, the discriminant validity is met.

Table 6. HTMT – confidence intervals bias corrected

	Original sample	Sample mean	Bias	2.5%	97.5%
PSK <-> COM	0.316	0.316	-0.001	0.153	0.469
QLT <-> COM	0.837	0.837	0.001	0.737	0.909
QLT <-> PSK	0.149	0.165	0.016	0.042	0.284
STRS <-> COM	0.909	0.910	0.002	0.832	0.974
STRS <-> PSK	0.254	0.261	0.007	0.116	0.403
STRS <-> QLT	0.879	0.880	0.001	0.792	0.949
TECH <-> COM	0.361	0.372	0.011	0.170	0.565
TECH <-> PSK	0.032	0.109	0.078	0.001	0.050
TECH <-> QLT	0.675	0.680	0.006	0.510	0.829
TECH <-> STRS	0.495	0.507	0.012	0.288	0.684

 

6. After reading the discussion and conclusions section, it is not clear what the conclusions and recommendations arising from the work. Therefore, the conclusions must be explained more clearly, and the recommendations and strategies to be carried out by the high education managers should be indicated.

Response: We added a paragraph under 6.1 subheading to better emphasize these recommendations and strategies:

Thus, in accordance with the results of our research, higher education managers should tailor their strategies to better fit the needs of both students and professors. As the findings show, communication problems and stress are the factors that influence mostly the quality of education. In conclusion, we recommend strategies focusing on improving communication between students and professors and finding solutions to reduce the stress that comes along with online education. For better communication, managers should invest in training programs for professors to better prepare them for providing interesting content to the students.

 

7. Likewise, due to the importance of the work, it is necessary to point out more clearly the limitations of the study, for which reason a subsection of conclusions should be developed.

Response: We better highlighted limitations in the last section of conclusions – 6.2 Limitations and future research directions.

            The limitations of our research refer to the fact that we conducted the quantitative analysis online, not face-to-face (which might have affected the dimension of the sample), due to geographical restrictions and reduced financial resources. Also, we developed a model with only five variables but in future research studies, we intend to add other constructs which might change the results, such as the influence of social groups, family support [49], personality traits, or physical and mental health.

Reviewer 2 Report

Comments to authors

The paper entitled: Online Education Management: A Multivariate Analysis of Students’ Perspective and the Challenges during Online Classes. The paper is well written and convenient for finding solutions for a better management of online education, starting from students’ perspective regarding the challenges they encountered in the last two years when online courses were imposed during the COVID-19 pandemic.

1. Introduction

Authors have mentioned that the topic of online education gained much attention especially in the last two years (2020-2022) because educational institutions had to move their courses online due to the
COVID-19 pandemic. However, is not shown updated studies which analysed in detail the new role of the online education system. Different research works will be appreciated to be quoted.

Authors explain that their research was not intended to reflect the challenges of the
pandemic, but to online education in general. Regarding to this issue it convenient to include a table with the evolution of online teaching whether it is introduced the number of students or schools that are currently following the online education.

Online education´s benefits and disadvantages are narrowed explained in the paper. Are different items that affect those ones that change drastically depending on the level of education, economy, technology, in different territories. Moreover, it is not introduced the socio-economic conditions in Romania, which is quite essential to understand how the online education system has evolved in the last years.

2. Literature Review

Authors mentioned that the novelty of the research consists in creating a model with variables that are considered the most challenging (communication problems, technical requirements, stress, professors’ skills) for the quality of online education. However, it is not well explained why authors have chosen those variables. It is not a matter of quoting authors, is simply to define the criteria of selecting those variables.

According to Coman et al. authors have mentioned that the lack of communication between professors and students in Romania was the least important during online classes. This statement needs to be explained more accurately.

Authors mentioned problems such as level of stress, psychological issues, fear, or decreased socialization. However there is no data added to the paper to show the relevance of those setbacks during Covid-19.

It is also said that that quality of online education is crucial. What methodology should they follow to improve the quality of that education? What kind of courses´ design?, how those feeling of being part of the community are created? And what kind of technological progress is necessary to provide that quality of education?

 

3. Research Methodology and Hypotheses Development

Indicators should be explained in the literature review. Table 1 should indicate the authors that support every indicator. Moreover, a brief explanation of every variable should be included in this section.

4. Results.

It is recommend using the HTMT criterion to assess discriminant validity. If the HTMT value is below 0.90, discriminant validity has been established between two reflective constructs.

Authors have confused Q-square is predictive relevance, with PLSpredict. The first one measures whether a model has predictive relevance or not (> 0 is good). Further, Q2 establishes the predictive relevance of the endogenous constructs. Q-square values above zero indicate that your values are well reconstructed and that the model has predictive relevance.

PLSPredict allow generating different out-of-sample and in-sample predictions (e.g., case-wise and average predictions), which facilitate the evaluation of the predictive performance when analyzing new data (that was not used to estimate the PLS path model). The analysis serves as a diagnostic for possible overfitting of the PLS path model to the training data.

Based on the procedures suggested by Shmueli et al. (2016), the current PLSpredict algorithm implementation in the SmartPLS software allows researchers to obtain k-fold cross-validated prediction errors and prediction error summaries statistics such as the root mean square error (RMSE), the mean absolute error (MAE), and the mean absolute percentage error (MAPE) to assess the predictive performance of their PLS path model for the manifest variables (MV or indicators) and the latent variables (LV or constructs). Note that all three criteria are available for the MV results, while it is only possible to compute the RMSE and MAE for the LV results. These criteria allow to compare the predictive performance of alternative PLS path models

5. Discussion and conclusions

In Discussion the impact of Coefficient of Determination should be also considered the relevance as well as the R2 divided into construct to study how every independent variable contribute the dependent variable.  

Author Response

Dear Reviewer,

Thank you for your observations and for the opportunity to improve our manuscript!

We are very grateful for taking the time to analyze the paper and make very useful, encouraging and thoughtful comments and recommendations.

We have read the evaluation and, based on the review reports, we performed revisions of our manuscript, as requested, highlighted with red into the manuscript.

 

1. Introduction

Authors have mentioned that the topic of online education gained much attention especially in the last two years (2020-2022) because educational institutions had to move their courses online due to the
COVID-19 pandemic. However, is not shown updated studies which analysed in detail the new role of the online education system. Different research works will be appreciated to be quoted.

Authors explain that their research was not intended to reflect the challenges of the
pandemic, but to online education in general. Regarding to this issue it convenient to include a table with the evolution of online teaching whether it is introduced the number of students or schools that are currently following the online education.

Online education´s benefits and disadvantages are narrowed explained in the paper. Are different items that affect those ones that change drastically depending on the level of education, economy, technology, in different territories. Moreover, it is not introduced the socio-economic conditions in Romania, which is quite essential to understand how the online education system has evolved in the last years.

Response: We added a few more references in the introduction [1-4, 12, 13]. Thus, the following text was added in the Introduction: Online education offered the opportunity of flexibility in a time when professors and stu-dents could not meet face to face [1]. Black et al. [2] consider “online education as an op-portunity equalizer”, offering access even to those in less developed regions, of course with the condition of having the technical infrastructure (Internet connection, devices for connecting online). Other recent studies present both the challenges and the opportunities provided by online education [3,4]. Watermeyer et al. [4] mention a “digital disruption” in UK universities determined by moving traditional education abruptly to an online format. As Adedoyin and Soykan [3] state, challenges should be “transformed to opportunities” for increased quality and efficiency.

Regarding the online education in Romanian universities, there are no statistics because only during the pandemic we used online education. After March 2022, it was not legal to have online classes in Romania because of a gap in legislation. This was possible only from October 2022 and only in hybrid form. No statistics for the moment because most universities started face to face and only in December and part of January, they will have online classes. This is related with the need to reduce electricity and heating bills.

We added the following text: Online education in Romania during the COVID-19 pandemic was possible because the country declared the state of emergency. Thus, between March 2020 and March 2022, online education was implemented in all higher education institutions. Still, after the sudden end of the state of emergency in March 2022, many universities faced a lack of legislation that did not allow them to continue the online classes. And this was difficult especially because students were not announced with enough time before about these changes. With the start of the new academic year (since October 2022), the country introduced the possibility for a hybrid form but still, with a prevalence of traditional format especially for seminars and labs. There are no statistics regarding the number of universities that use the hybrid form but many important universities implemented this form especially during winter months.

Regarding the perspective of advantages and disadvantages of online education taking into account the specific conditions in Romania, we added some studies conducted by professors and students in the most important universities in the country which reveal the advantages and the disadvantages during the two years. Thus, we added the following text as well as the additional references [12,13]:

The benefits and the disadvantages of online education [7-10] are influenced by many factors. Thus, if the Internet connection is good, online education is seen as an advantage, meanwhile if the broadband coverage is a problem, then online education is seen as a disadvantage or at least a challenge to work on. According to World Population Review [11], the speed of Internet connection is one of the fastest in the world which constitutes an important foundation for online education. As we previously mentioned, the evolution of online education after the state of emergency ended in March 2022, Romanian universities could not continue online classes because of a gap in legislation which was corrected only for the start of academic year in October 2022. Potra et al. [12] conducted research on students in first year of their studies and the problems revealed were: “information overload, limited interaction, teacher-related hindrances and presence and concentration hurdles”. Another report [13] reveals the conclusions of students from the most import important universities in Romania: digital competencies were not a problem, technical difficulties were not significant, the access to Internet was not limited, computers were performant enough, the digital resources were available in high proportion, the time was not a problem. Lack of motivation was a problem for half of the students. The study also mentions the perception of professors that students with good academic results performed well also during online classes but for students with low academic results, the problems in learning increased as well as the gap between students. 

2. Literature Review

Authors mentioned that the novelty of the research consists in creating a model with variables that are considered the most challenging (communication problems, technical requirements, stress, professors’ skills) for the quality of online education. However, it is not well explained why authors have chosen those variables. It is not a matter of quoting authors, is simply to define the criteria of selecting those variables.

According to Coman et al. authors have mentioned that the lack of communication between professors and students in Romania was the least important during online classes. This statement needs to be explained more accurately.

Authors mentioned problems such as level of stress, psychological issues, fear, or decreased socialization. However there is no data added to the paper to show the relevance of those setbacks during Covid-19.

It is also said that that quality of online education is crucial. What methodology should they follow to improve the quality of that education? What kind of courses´ design?, how those feeling of being part of the community are created? And what kind of technological progress is necessary to provide that quality of education?

 Response: Regarding the variables we chose, we considered the national report of student in Romanian universities [13] and extracted the mot challenging variables from there, Of course, there can be other factors to be considered and we also mentioned them in Section 6.2 Limitations and future research direction. We added the report [13] to explain our choice. Of course, all the sources in the literature review are important as a foundation for our research.

We added a better explanation for Coman study: The most important problems were technical issues and lack of technical skills [20]. The results are normal for the time of the study (second semester, the beginning of online classes in Romania) taking into account that neither professors, nor students were prepared to move to a complete online format. In this context, students saw technical problems as having a higher importance than the interaction with their professors.

Regarding the psychological issues (fear, stress, isolation), we added in the text the references we used [4,8,9,12,13]. The stress variable is also detailed in subsection 2.4.

Regarding the aspects mentioned on quality education in subsection 2.3, we added a few recommendations for managers in higher education institutions in order to raise the quality level, mainly trainings for professors to help them with communication and also with offering a more interesting content for students. We added these conclusions and recommendations in subsection 6.1 Theoretical and practical implications:

Thus, in accordance with the results of our research, higher education managers should tailor their strategies to better fit the needs of both students and professors. As the findings show, communication problems and stress are the factors that influence mostly the quality of education. In conclusion, we recommend strategies focusing on improving communication between students and professors and finding solutions to reduce the stress that comes along with online education. For better communication, managers should invest in training programs for professors to better prepare them for providing interesting content to the students.

3. ResearchMethodology and Hypotheses Development

Indicators should be explained in the literature review. Table 1 should indicate the authors that support every indicator. Moreover, a brief explanation of every variable should be included in this section.

Response: Thank you for the recommendation. Because the other reviewers also suggested it, we added the references we used for the indicators of the model in table 1, the last column. All constructs are referenced in the literature review section and now, we have the correlation clearer in table 1 by adding the references we presented in the literature review section for each of the constructs in the model.

All five variables in the model are explained in Section 2. Literature review but we added a brief explanation of the hypotheses in order to make them more visible in relation to figures 1 and 2 (now 2 and 3 because another reviewer asked us to add a new figure before them). Thus, we added the following text in Section 3: Figure 2 illustrates the model we proposed for the present research (H1: COM -> QLT; H2: COM -> STRS; H3: PSK -> COM; H4: PSK -> QLT; H5: PSK -> STRS; H6: STRS -> QLT; H7: TECH -> COM; H8: TECH -> QLT; H9: TECH -> STRS) and Table 1 details the model’s constructs, items and codes.

4. Results.

It is recommend using the HTMT criterion to assess discriminant validity. If the HTMT value is below 0.90, discriminant validity has been established between two reflective constructs.

Response: Thank you for the recommendation. We added, as the other reviewers also requested it, the calculation for HTMT confidence intervals bias corrected. Thus, we added the Table no 6 and the explanations. Also, we added two more references for explaining the results. The text added is the following:

For checking the discriminant validity of the model, we calculated Heterotrait-monotrait ratio (HTMT) confidence intervals bias corrected in table 6, as suggested by Ringle [43]. The confidence intervals in table 6 show a good discriminant validity between the constructs even if the correlation between STRS and COM has a borderline value for HTMT (0.909). As Ringle [43] and Henseler et al. [44] state, if the confidence intervals for HTMT do not include the value 1, the discriminant validity is met.

                             Table 6. HTMT – confidence intervals bias corrected

	Original sample	Sample mean	Bias	2.5%	97.5%
PSK <-> COM	0.316	0.316	-0.001	0.153	0.469
QLT <-> COM	0.837	0.837	0.001	0.737	0.909
QLT <-> PSK	0.149	0.165	0.016	0.042	0.284
STRS <-> COM	0.909	0.910	0.002	0.832	0.974
STRS <-> PSK	0.254	0.261	0.007	0.116	0.403
STRS <-> QLT	0.879	0.880	0.001	0.792	0.949
TECH <-> COM	0.361	0.372	0.011	0.170	0.565
TECH <-> PSK	0.032	0.109	0.078	0.001	0.050
TECH <-> QLT	0.675	0.680	0.006	0.510	0.829
TECH <-> STRS	0.495	0.507	0.012	0.288	0.684

Authors have confused Q-square is predictive relevance, with PLSpredict. The first one measures whether a model has predictive relevance or not (> 0 is good). Further, Q2 establishes the predictive relevance of the endogenous constructs. Q-square values above zero indicate that your values are well reconstructed and that the model has predictive relevance.

PLSPredict allow generating different out-of-sample and in-sample predictions (e.g., case-wise and average predictions), which facilitate the evaluation of the predictive performance when analyzing new data (that was not used to estimate the PLS path model). The analysis serves as a diagnostic for possible overfitting of the PLS path model to the training data.

Based on the procedures suggested by Shmueli et al. (2016), the current PLSpredict algorithm implementation in the SmartPLS software allows researchers to obtain k-fold cross-validated prediction errors and prediction error summaries statistics such as the root mean square error (RMSE), the mean absolute error (MAE), and the mean absolute percentage error (MAPE) to assess the predictive performance of their PLS path model for the manifest variables (MV or indicators) and the latent variables (LV or constructs). Note that all three criteria are available for the MV results, while it is only possible to compute the RMSE and MAE for the LV results. These criteria allow to compare the predictive performance of alternative PLS path models

Response: Thank you for the observation. I changed the title of the table to be more accurate and also added in the text how we calculated Q²predict. In the current version of SmartPLS, version 4 which we used, we calculated Q²predict by applying PLSpredict which I think in the previous versions of the software were not overlapped.

Thus, we changed the text to better explain the fact that I used the values for Q2predict by applying PLSpredict in version 4 of the software. I think it was different in versions 2 and 3. The changes in the text are the following:

In order to determine if the model in our research has a predictive relevance, we calculated Q²predict in table 8. Because the Q² predict values for the dependent constructs (COM, QLT and STRS) are higher than 0, we can state that the model we proposed has a high predictive relevance. Q²predict is determined by applying PLSpredict in SmartPLS.

Table 8. Q²predict (LV prediction summary in PLSpredict results)

Constructs	Q2predict
COM	0.121
PSK	-
QLT	0.217
STRS	0.144
TECH	-

Source: Own work using SmartPLS, version 4

5. Discussion and conclusions

In Discussion the impact of Coefficient of Determination should be also considered the relevance as well as the R2 divided into construct to study how every independent variable contribute the dependent variable.  

Response: Thank you for the suggestion. We added a paragraph to better emphasize the importance of R square and how these results can be used by managers in establishing their strategies in universities.

The paragraph added is the following: Figure 3 shows the coefficients of determination measured by R square in SmartPLS. Thus, the high R2 for QLT (0.691) show the important influence played by STRS, COM. TECH and PSK with the highest impact from STRS and COM (expressed by H1 and H6). This is relevant because decision factors like the managers in the higher education system could use these results for raising the quality level of online education in higher education institutions. The second most important coefficient of determination is of STRS (0.605) which is also mostly influenced by COM. Because communication problems raise the lev-el of stress felt by students, managers should develop training programs for professors in order to reduce the communication gap between students and their professors.

 

 

Reviewer 3 Report

The article is proposed to be supplemented with a flowchart illustrating the research technique in the section on methods.

The article holds promise and is well-organized - minor editing is needed.

 

Author Response

Dear Reviewer,

Thank you for your observations and for the opportunity to improve our manuscript!

We are very grateful for taking the time to analyze the paper and make very useful, encouraging and thoughtful comments and recommendations.

We have read the evaluation and, based on the review reports, we performed revisions of our manuscript, as requested, highlighted with red into the manuscript.

 

1. The article is proposed to be supplemented with a flowchart illustrating the research technique in the section on methods.

Response: Thank you for the recommendation. We added the flowchart in section 3.

2. The article holds promise and is well-organized - minor editing is needed.

Response: We proofread the article and made some minor corrections in the text, mostly typo ones.   

Reviewer 4 Report

This is a very sound and well written article, very well structured and presented. In my opinion, the literature review could be wider and include some points that were mentioned in the discussion, for example. The results are clear, and the discussion presents a clear argumentation. All in all, this is an article with no evident issues and so I’m happy to recommend it for publication.

Best wishes

Author Response

Thank you for your appreciation. We added a few more references in the literature review. Thank you.

The following text was added in the literature review in red:

The relationship between stress during online learning and the quality of education is also highlighted by Altaf et al. [30] which conclude that for medical students, the online experience was less stressful than face-to-face learning which might be explained by the inherent stress specific to this profession.

 Cullinan et al. [37] also address the problem of disparities between students with access to Internet and students from regions with less coverage and the way these problems affect the quality of education if universities do not provide support for these latter students.

The references are also highlighted in the reference list.

Altaf, R.; Kling, M.; Hough, A.; Baig, J.; Ball, A.; Goldstein, J.; Brunworth, J.; Chau, C.; Dybas, M.; Jacobs, R.J.; Costin, J. The Association Between Distance Learning, Stress Level, and Perceived Quality of Education in Medical Students After Transitioning to a Fully Online Platform. Cureus 2022, 14(4), e24071. https://doi.org/10.7759/cureus.24071

Cullinan, J., Flannery, D., Harold, J., Lyons, S. and Palcic, D., 2021. The disconnected: COVID-19 and disparities in access to quality broadband for higher education students. International Journal of Educational Technology in Higher Education, 18(1), pp.1-21.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Dear authors,

I agree with the changes inserted in the manuscript. However, Table 8. shoul be changed, instead of "Q2predict (LV prediction summary in PLSpredict results)" should say: "Cross-validated redundancy".

On the one hand, according to (Rigdon, 2014; Sarstedt et al., 2014) the Q2 is a means for assessing the inner model's predictive relevance. The measure builds on a sample re-use technique, which omits a part of the data matrix, estimates the model parameters and predicts the omitted part using the estimates. The smaller the difference between predicted and original values the greater the Q2 and thus the model's predictive accuracy. Specifically, a Q2 value larger than zero for a particular endogenous construct indicates the path model's predictive relevance for this particular construct. It should, however, be noted that while comparing the Q2 value to zero is indicative of whether an endogenous construct can be predicted, it does not say anything about the quality of the prediction.

On the other hand, hShmueli et al. (2016) proposes a set of procedures for prediction with PLS path models and the evaluation of their predictive performance. They allow generating different out-of-sample and in-sample predictions (e.g., case-wise and average predictions), which facilitate the evaluation of the predictive performance when analyzing new data (that was not used to estimate the PLS path model). The analysis serves as a diagnostic for possible overfitting of the PLS path model to the training data. Based on the procedures suggested by Shmueli et al. (2016), the current PLSpredict algorithm implementation in the SmartPLS software allows researchers to obtain k-fold cross-validated prediction errors and prediction error summaries statistics such as the root mean square error (RMSE), the mean absolute error (MAE), and the mean absolute percentage error (MAPE) to assess the predictive performance of their PLS path model for the manifest variables (MV or indicators) and the latent variables (LV or constructs). Note that all three criteria are available for the MV results, while it is only possible to compute the RMSE and MAE for the LV results. These criteria allow to compare the predictive performance of alternative PLS path models. Sharma et al.’s (2019) Monte Carlo simulation shows that the RMSE and mean absolute deviation MAE are particularly suitable when the aim is to select the best predictive model among a set of competing models. Researchers need to compare RMSE and MAD values for alternative model set-ups and select the model, which minimizes RMSE and MAD values in the latent variable scores

Author Response

First of all, we really thank you for the explanations regarding Q2 predict and for the references you provided. Indeed, we had a problem of how to caption Table 8 and now, according to your suggestion, it is recaptioned as Cross-validated redundancy, being more in line with what it represents.

We also enriched the literature review with a few more references. References 30 and 37 in the current form of the manuscript.

Kind regards,

The authors

Author Response File: Author Response.pdf