Analyzing Takeaway E-Bikers’ Risky Riding Behaviors and Formation Mechanism at Urban Intersections with the Structural Equation Model

Round 1

Reviewer 1 Report

Full Title: Analyzing Takeaway E-bikers’ Risky Riding Behaviors and Formation Mechanism at Urban Intersections with the Structural Equation Model

Objective: The present study aimed to investigate relationships among five variables (individual characteristics, safety attitude, riding confidence, risk perception, and risky riding behavior) and the risky riding behaviors of takeaway e-bikers. Although the topic is interesting, the manuscript needs more work and effort in each section:

1.      The manuscript reviews the existing literature in the literature review section; however, it needs another section to introduce the research model and hypotheses development. In other words, a specific sub-section for hypotheses is needed. The authors must connect their theoretical background and their research design and model. Specifically, identify more literature that is (more) relevant for their research constructs.

2. A scale of safety attitude, riding confidence, risk perception, and risky riding behavior for takeaway e-bikes was developed by researchers. The risky riding behavior scale questionnaire for the takeaway e-bike contains five dimensions: individual characteristics, safety attitude, riding confidence, risk perception, and risky riding behavior. The instrument is formed by six demographic variables, such as sex, age, and income along with 13 Likert type items. Since this is a scale development study or the author(s) developed a scale through the study, first an Exploratory Factor Analysis (EFA) should be conducted in study 1.

a.   Pattern Matrix or Rotated Component Matrix should be provided.

b.   Commonality values should be reported.

c.   Descriptive Statistics (Mean, S.D., Skewness, Kurtosis) and Reliability coefficients for each subdimension should be reported.

3. Next, a Confirmatory Factor Analysis (CFA) should be conducted in study 2.

a.   Study 2 should be conducted by using a different data set.

b.   Reliability of the subscales should be reported.

c.   Convergent and discriminant validity should be tested.

d.   CR and AVE values along with HTMT results should be reported to evaluate the convergent and discriminant validity.

e.   The correlations matrix should be provided.

f.    The CFA should be conducted to test the measurement model. Model fit estimates including χ2/DF, GFI, AGFI, CFI, TLI, IFI, RMSEA [LO90, HI90], SRMR should be reported.

4. Since this is a cross-sectional study, common method bias may arise from certain tendencies that respondents apply. Therefore, common method bias should be an important concern and assessed.

5. In the proposed model demographic variables are categorical, whereas Likert-type items are continuous. For demographic variables, which are categorical in nature, regression analysis may not be suitable. Multi-group analysis can be used to test moderation by categorical variables.

6. Figures should be in high resolution (TIFF images with a minimum of 300 dpi).

7. Discussion and Conclusion should be revised, accordingly. To address the issues related to scale development, you may refer to:  https://doi.org/10.1177/0266666921997512, and https://doi.org/10.1016/j.paid.2020.110108.

Proof reading is required. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This study investigated a risky riding behavior using data collected from questionnaires. A SEM model is used to evaluate the internal potential relationships between influencing factors and risky riding behaviors, which in order to study the internal formation mechanisms of risky riding behaviors of takeaway e-bikers at urban intersections and propose corresponding improvement strategies. The results showed that individual characteristics, safety attitude, riding confidence and risk perception are significantly related to risky riding behavior. Overall, the paper is written in a good manner. The following comments may be helpful for further improvement.

In the background, a brief summary about what this study has done in terms of methodology and research contents should be provided, besides research questions.

While I appreciate the vast literature reported, its discussion is not well organized. At the moment, it looks like an unstructured description of current papers, listed just one after the other. It needs to be reorganized around the specific objective of this paper with the aim to clarify the choices made in this paper and why these are relevant.

SEM could provide some insights about the effects of different subjective factors to some extent. However, other objective factors such as environmental factors also affect the risky driving behavior. It would be an interesting direction to use hybrid choice modeling to quantitatively model both subjective and objective factors. This is recommended to be added in limitation statement after conclusions. Two relevant reference is suggested for you information

Revealing psychological inertia in mode shift behavior and its quantitative influences on commuting trips

Young drivers' night-time mobility preferences and attitude toward alcohol consumption: A hybrid choice model

English should be polished again before publication

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The article raises a significant problem related to road safety. In my country, many food deliverers travel on bicycles. Their behavior on the road is different from that of cyclists in general.

It is not indicated whether the measurement results presented in section 3.1. they concern one day or many days in the indicated periods. If these are data from one day, they are appalling! In my country, there are visible offenses of food suppliers on bicycles, but not so many.

What are the technical requirements for e-bikes in the research area? In my country, an e-bike can have a maximum power of 250 W and assist the cyclist while pedaling only up to 25 km/h. In Figure 1b, I see that the bicycle has no pedals. Such vehicles in my country are classified as a moped.

How was vehicle speed tested in the field? What values are considered abnormal behavior?

Were the examined intersections different from each other? Please include the phrase "Analysis of cyclists' behavior on different infrastructure elements" in the literature analysis - you can find some articles on this subject and refer to this issue in the summary.

Table 8 is truncated; please present it properly to make it readable for readers.

Why is Cramer's V not used as a measure in Table 8? I do not require a change in the article, but I ask for clarification in response to the review.

Technical Notes:
In many places in the article, the titles of chapters and tables are not on the same page as the actual content, e.g., lines 193, 386.
Line 167-168, the wording "risky driving behaviors at intersections at home and abroad." is incomprehensible.
There are incorrect table formats, e.g., at the bottom of page 11, and wrong transpositions of words in tables (e.g., at the top of page 15).

I find the article valuable and recommend it for publication. The conclusions regarding the education of e-bike riders are correct.

Good luck! I hope to read the revised version of the article soon.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

While the authors have made commendable efforts to address several of the concerns I raised, it's important to highlight a few remaining limitations that warrant further attention.

This study employs a cross-sectional design, and it's important to acknowledge the potential emergence of common method bias due to specific response tendencies. Consequently, addressing common method bias becomes a crucial consideration that requires thorough evaluation. It's noteworthy that while the revised version appropriately focuses on tests for validity and reliability, it's equally vital to emphasize the assessment of common method bias as a distinct concern.

In the proposed model, it's notable that demographic variables take on a categorical nature, while Likert-type items remain continuous. Given this distinction, it's prudent to acknowledge that regression analysis might not be the most suitable approach for categorical demographic variables. To address this, I've recommended the utilization of an alternative method, specifically designed to examine moderation involving categorical variables. Notably, while the authors have indeed conducted a multi-group analysis, an oversight exists in the update of Figures 3 and 4 to accurately reflect the moderating influence of demographic variables.

Lastly, it's worth considering the inclusion of pertinent references for the enhancement of your work. These scale development papers could offer valuable insights: https://doi.org/10.1177/0266666921997512, and https://doi.org/10.1016/j.paid.2020.110108.

Proof reading is required.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf