Next Article in Journal
Post-Recovery Mental Health Outcomes in Young Adults with Mild-to-Moderate COVID-19: A Cross-Sectional Study
Previous Article in Journal
Forecasting COVID-19 Cases, Hospital Admissions, and Deaths Based on Wastewater SARS-CoV-2 Surveillance Using Gaussian Copula Time Series Marginal Regression Model
 
 
Article
Peer-Review Record

COVID Stress Factors Affecting Remote Work Acceptance

by Cheong Kim
Reviewer 1: Anonymous
Reviewer 2:
Reviewer 3: Anonymous
Submission received: 5 January 2025 / Revised: 12 February 2025 / Accepted: 17 February 2025 / Published: 18 February 2025
(This article belongs to the Section COVID Public Health and Epidemiology)

Round 1

Reviewer 1 Report

In this paper, the author uses Probabilistic Structural Equation Modeling (PSEM) to analyze COVID-related stress factors affecting work acceptance. The manuscript is well-written, with a detailed introduction and discussion.

Major Comments:

  1. In the Results section, Table 3 presents the performance of EQ compared to other machine learning models. However, it is unclear whether this refers to the EQ algorithm or the Bayesian EQ algorithm. Additionally, compared to other machine learning models, the EQ algorithm does not appear to demonstrate a significant performance advantage. Could the author clarify?
  2. The clustering of variables in the Results section is also unclear. The author states that the "machine learning-based Bayesian PSEM in this research suggests that some of the items in the contamination construct could be included in the danger construct." A new construct was then built. If so, there should be two different clustering schemes, yet in the subsequent analysis, only the clustering in Figure 2 seems to be considered. Could the author clarify this inconsistency?
  3. In the Results section, the author states, "The only child node of acceptance is compulsive checking, while socio-economic consequences is a child node of compulsive checking, and both compulsive checking and socio-economic consequences have statistically significant effects on the target node acceptance." It seems that the child node structure directly influences the effects on the target node (acceptance). Could the author further elaborate on this relationship?
  4. The statement, "Considering that the dependent variable was coded as a three-class node, the test score seems to be acceptable," appears to overstate the model's accuracy, as the actual accuracy is quite low. This statement does not seem to sufficiently validate the model's usefulness. Could the author provide additional justification or reconsider this claim?

 

Minor comments:

  1. In Table 2, there is no need to add % after the values.
  2. The capitalization of "Equivalence Classes Algorithm" should be consistent throughout the manuscript.

Author Response

I thank the reviewer for the timely and detailed feedback on the manuscript, and I believe that the feedback was much beneficial for us to improve the research significantly. I have prepared a revised version that reflects the suggestions by the reviewers. I further addressed the reviewers’ comments one by one; a precise discussion of every feedback point can be found in my answers. I hope that the new version of the manuscript has satisfied your concerns, was enhanced enough to contribute to academia and the related field, and will help more future readers understand the research.

 

(Comment 1) In the Results section, Table 3 presents the performance of EQ compared to other machine learning models. However, it is unclear whether this refers to the EQ algorithm or the Bayesian EQ algorithm. Additionally, compared to other machine learning models, the EQ algorithm does not appear to demonstrate a significant performance advantage. Could the author clarify?

(Response 1) I sincerely appreciate the reviewer's productive comments. To clarify terminology, EQ has been revised to Bayesian EQ. Additionally, as the reviewer pointed out, Bayesian EQ does not demonstrate a significantly superior performance compared to other machine learning models. However, in this study, I aimed to show that Bayesian EQ exhibits comparable performance to other machine learning models, as presented in Table 3. Moreover, due to its advantages, such as visualization, I chose Bayesian EQ as the methodology for this study. To enhance clarity, I have revised the corresponding content in the manuscript, which can be found on page 6, line 198.

 

(Comment 2) The clustering of variables in the Results section is also unclear. The author states that the "machine learning-based Bayesian PSEM in this research suggests that some of the items in the contamination construct could be included in the danger construct." A new construct was then built. If so, there should be two different clustering schemes, yet in the subsequent analysis, only the clustering in Figure 2 seems to be considered. Could the author clarify this inconsistency?

(Response 2) I acknowledge that the explanation in the manuscript was not sufficiently clear, as the reviewer pointed out. In the process of constructing the PSEM model, variable clustering assigns the same color to nodes with similar internal consistency, and these nodes later form new factors through multiple clustering. Therefore, in this study, some nodes from Danger and Contamination form one group (same color), while the remaining nodes from Contamination form another group. To clarify this, I have added further explanation to the manuscript, which can be found on page 7, line 213.

 

(Comment 3) In the Results section, the author states, "The only child node of acceptance is compulsive checking, while socio-economic consequences is a child node of compulsive checking, and both compulsive checking and socio-economic consequences have statistically significant effects on the target node acceptance." It seems that the child node structure directly influences the effects on the target node (acceptance). Could the author further elaborate on this relationship?

(Response 3)  I acknowledge that the explanation was insufficient, as the reviewer pointed out. To clearly describe the relationship between the target node and its child nodes, I have revised the manuscript accordingly. This revision can be found on page 9, line 241.

 

(Comment 4) The statement, "Considering that the dependent variable was coded as a three-class node, the test score seems to be acceptable," appears to overstate the model's accuracy, as the actual accuracy is quite low. This statement does not seem to sufficiently validate the model's usefulness. Could the author provide additional justification or reconsider this claim?

(Response 4) I fully agree with the reviewer's comments. I have revised the manuscript to explicitly state that, given the sample size limitation (568 samples), the performance metrics—except for the ROC value—are barely acceptable. Additionally, I have emphasized that the model in this study serves as an initial baseline model. This revision can be found on page 6, line 192.

 

(Comment 5) In Table 2, there is no need to add % after the values.

(Response 5) All % in Table 2 is deleted.

 

(Comment 6) The capitalization of "Equivalence Classes Algorithm" should be consistent throughout the manuscript.

(Response 6) Now, "Equivalence Classes Algorithm" is consistent throughout the manuscript.

Reviewer 2 Report

This paper is generally well-written and covers a highly interesting topic. I would like to suggest a few minor comments that may help improve the quality of the manuscript:

  1. Provide a clearer explanation of why PSEM is a more appropriate method than traditional SEM.
  2. There are a few typos and grammatical errors in the document (e.g., Beysian networkduing, etc.), so it would be good to correct them.
  3. Consider adding a few more recent references relevant to this field.

I hope these suggestions help enhance the quality of the paper.

Author Response

I thank the reviewer for the timely and detailed feedback on the manuscript, and I believe that the feedback was much beneficial for us to improve the research significantly. I have prepared a revised version that reflects the suggestions by the reviewers. I further addressed the reviewers’ comments one by one; a precise discussion of every feedback point can be found in my answers. I hope that the new version of the manuscript has satisfied your concerns, was enhanced enough to contribute to academia and the related field, and will help more future readers understand the research.

 

(Comment 1) Provide a clearer explanation of why PSEM is a more appropriate method than traditional SEM.

(Response 1) I acknowledge that the explanation was insufficient, as the reviewer pointed out. I have revised the content to provide a clearer explanation of why PSEM is a more appropriate method than traditional SEM. Please kindly see page 5, line 147.

 

(Comment 2) There are a few typos and grammatical errors in the document (e.g., Beysian network, duing, etc.), so it would be good to correct them.

(Response 2) All typos and grammatical errors have been corrected.

 

(Comment 3) Consider adding a few more recent references relevant to this field.

(Response 3) I have added more than 10 recent references relevant to this research.

Reviewer 3 Report

This paper is quite interesting, but I have a few recommendations:

1) Please include the F1 score in Table 3.
2) PSEM is not a well-known approach; therefore, please provide a more detailed discussion on PSEM for the readers.

This paper is quite interesting, but I have a few recommendations:

1) "Cont-danger" should also be explained in the abbreviation section.

Author Response

I thank the reviewer for the timely and detailed feedback on the manuscript, and I believe that the feedback was much beneficial for us to improve the research significantly. I have prepared a revised version that reflects the suggestions by the reviewers. I further addressed the reviewers’ comments one by one; a precise discussion of every feedback point can be found in my answers. I hope that the new version of the manuscript has satisfied your concerns, was enhanced enough to contribute to academia and the related field, and will help more future readers understand the research.

(Comment 1) Please include the F1 score in Table 3.

(Response 1) The F1 score is now included in Table 3.

 

(Comment 2) PSEM is not a well-known approach; therefore, please provide a more detailed discussion on PSEM for the readers.

(Response 2) I acknowledge that PSEM could be new to future readers, as the reviewer pointed out. I have revised the content to provide a more detailed discussion of PSEM. Please kindly see page 5, line 147.

 

(Comment 3) "Cont-danger" should also be explained in the abbreviation section.

(Response 3) "Cont-danger" is now explained in the abbreviation section.

 

 

Back to TopTop