Smartphone GPS Locations of Students’ Movements to and from Campus

Round 1

Reviewer 1 Report

The paper assesses the student engagement on a Canadian campus, using their GPS location history.

Overall, the manuscript is well-written. However, there are some concerns about the significance of results and the research contribution, which I will outline below.

1) The general conclusions were that a) Students spent more hours of the day on campus when they lived further from campus, and if they had a relatively high-speed commute, b) Students with slow commutes tended to come into campus fewer days of the week, and even less frequently if they lived farther from campus, c) Visit frequency is more constrained by a students’ timetable than the duration of their visit and is less responsive to external features such as semester, or more fine-scale commute characteristics.

which are unsurprising, and have been previously reported elsewhere. Therefore, I struggle to see the significance of the results.

Perhaps, publishing the dataset (without anonymised user information) would somewhat enhance the contribution of the paper.

Additionally, more visualisations and detailed analysis of the dataset would compensate for the lack of originality.

2) Figure 1 is ambiguous, and does not demonstrate the scale of the travelling distance. It would be clearer to provide some measuring unit for the routes, as well as some landmarks (e.g., station name).

3) What is the unit of the length of commute in Figure 4 ?

4) The Related work section is missing.
Since this is a popular topic, it would be beneficial to the readers to outline other approaches in measuring student engagement, as well as comparing the results with other work using GPS data.

Author Response

Thank you to the reviewer for their due diligence in reviewing our manuscript. We have included our responses and made the necessary changes to the manuscript

1) Unfortunately, we do not have ethical approval for sharing the database. See our answer to question 4 re: originality

2) Thank you, we agree that Figure 1 is not clear and we have added North Arrow, scale bar, and some road labels for context related to the City of Calgary.

3) We thank the reviewer for identifying an error on our part with Figure 4 X-axis. It is not log-transformed, rather it was still on its scaled axis. We have rectified this error by making the X-axis the same as Figure 3 and correcting the figure caption.

4) We, approached measuring and quantifying campus engagement thru GPS location data, i.e., the time a student spends outside lectures on campus, because of the large student commuter population in many Canadian campuses’ and because the validity of previous data yielded by self-report measures has been questioned (Assor & Connell, 1992). This so-called study at home or ‘stayeducation’ (Pokorny et al., 2017), refers to commuter students and the distance and travel experienced by them. We chose the use of location data (which arguably could be considered novel and original given the lack of research in this area on this topic), not surveys as many others have done (Boulton et al., 2019). Surveys inherently have bias and it is difficult to quantify how long (duration) students remain on campus.

Reviewer 2 Report

This article is related to GPS outdoor positioning system. The research on student's movement trajectory and campus life is very interesting. However, the research methods and explanations have several shortcomings. Several issues must be clarified and improved in order to correctly describe the system and be able to reproduce the system through other studies.

1. This work uses a Bayesian regression model to study the correlation between the length and frequency of campus visits and commuting characteristics. Did the author compare and analyze the research through other models?
2. Please explain the GPS trajectory in more detail. At the same time, will there be data loss due to the loss or weakening of the GPS signal of the motion track? How to deal with such data?
3. “To assess student engagement on campus, we collected smartphone GPS location histories from volunteers (n=280) attending university in a major  Canadian city.” and "We recruited 280 University of Calgary students over a two-year period starting in April 2015" Please explain how to select a sample of students. For a city, is there a problem of insufficient data for n=280?
4. I think the author can add more details about Section 3. Only "Visit length" and "Visit frequency" cannot show convincing conclusions. Is it possible to add some simulations about other parameters?
5. The current indoor positioning technology has developed very well. I think this work can add the movement trajectory of indoor positioning and the length of stay in various buildings. Through the study of these parameters, it is possible to analyze in more detail the potential of students to participate in campus life.

Looking forward to the author's reply. Best regards

Author Response

Thank you to the reviewer for their due diligence and their insightful questions related to reviewing our manuscript. We have included our responses and made the necessary changes to the manuscript.

1) Thank you for this question. We investigated a number of different models in terms of what we felt best represented the response variables and decided that a beta distribution and ordered logit were the most appropriate model structures for the data. We constructed multiple models that represented competing hypotheses to try and ascertain factors influencing students' time spent on campus and frequency of visits.

2) Thank you for the opportunity to clarify this. We have added two sentences to try and further explain how we created trips (by combining steps that were straight-line linear features between consecutive GPS relocations). We mention on line 145 that we removed locations that had poor accuracy to avoid inaccurate representation of the commute trips. Please note there is a thorough explanation of the steps we took to “clean” the data in Mobile phone GPS data and Identifying residence-campus commutes sections of the methods.

3) In our initial recruitment of volunteer students in year 1, we reached 127 university students (234,709 hours of behavioural observation; Ladle et al., 2018; Galpern et al., 2018) and this was followed by a second recruitment year 2 and we hit a kind of saturation at 280 students. We used many recruitment strategies, including student-to-student recruitment, incentivization by gift card, during both fall and winter semesters and in locations where large numbers of students pass through, etc. We did not do a sample size calculation, however student research on university students using cross-sectional surveys with two main outcome variables shows that N=270-290 would have enough power with adequate representation of the sample population. Our goal was to measure and quantify student campus engagement thru GPS location data. Therefore, we do not need to be representative of the city, give our population is based on commuter students at university campuses.

4 and 5) Thank you for the opportunity to clarify these details. We chose these two metrics as they were quantifiable from solely GPS data, which we believe is a strength of this manuscript. These two variables (“Visit length" and "Visit frequency”) are key parameters in understanding student engagement with university campuses. More detailed, finer-scale understandings of student engagement, for example, what they did with their time on campus, etc., would require data, likely a combination of methods, beyond the scope of GPS data, e.g. survey data and/or qualitative interview data. Our manuscript aims to show the information that we can garner from solely using smartphone GPS data. We, therefore, believe that this approach and these response variables do show convincing conclusions.

Reviewer 3 Report

This article investigated the relationship of students’ campus engagement and characteristics of commute by using GPS data. The conclusions drawn in article have a certain meaning for Canadian university management to make decisions. However, some places need to be supplemented.

1. The sample number is 280. Please explain the choice of sample size and whether the sample size can characterize the group characteristics.
2. In line 30, the expression of “a sizeable portion” is not rigorous.
3. In line 152, the reason using time between 2200 and 0500 hours needs to be explained.
4. Some formulas are missing for a more rigorous expression.

Author Response

Thank you to the reviewer for your due diligence in reviewing our manuscript and the opportunity to clarify. We have included our responses and made the necessary changes to the manuscript.

1) In our initial recruitment of volunteer students, we reached 127 university students (234,709 hours of behavioural observation; Ladle et al., 2018; Galpern et al., 2018) and this was followed by second recruitment, however, we hit saturation at 280 students. We did not do a sample size calculation, however, research on university students using cross-sectional surveys with two main outcome variables shows that N=270-290 would have enough power with adequate representation of the sample population.

2) We have changed this term 'sizeable' to 'one third or more ' based on Canadian research.  

3) Thank you, we have added a few words for clarity. Under the heading: Identifying residence-campus commutes, (L156) the first line reads "We defined a student’s primary place of residence as the most used location between 2200 and 0500 hours."

We assumed on average, that most students would spend the hours between 2200 and 0500, at their place of residence sleeping, and not undertaking activities outside of their residence.

4) With due respect, we are not sure what the formulas are missing.

Reviewer 4 Report

The paper presents a students' engagement assessment from a dataset built from a smartphone GPS locations data, collected and filtered according to some parameters such as spatial accuracy, relocation frequency and trip duration. The engagement is characterized by visit length and visit frequency, metrics that are extract from that dataset. It is well written, clearly and objectively.

I have only some few comments.

- I suggest to make the title more specific, for instance, including the 'engagement'. (Smartphone, instead of smart phone, is most used in the text)

- I suggest to write 'gps location data' instead of only 'gps data', along the whole text. (location meaning coordinates and time)

- In the line 201, 'Between 2012 and 2017' is written, but in the line 139, it is stated that students were recruited starting in April 2015. Would you please check it?

- In the Figure 4, identifications (a) and (b) are missing. Also, Figure S4a and Figure S4b are mentioned in the text; lines 260 and 263.

- Were the limits for spatial accuracy (100m) and trip's distance (10Km) and duration (2h) set by analyzing the GPS location data? I missed this information in the text.

Author Response

Thank you to the reviewer for their suggestions as it does strengthen the manuscript quality. 

1Thank you. We have accepted this suggestion – Smartphone.

2) Thank you. We have accepted this suggestion for our research and made the change to 'GPS location data.'

3) Thank you for finding that typo error of 2012. We have made the change to 2015.

4) Thank you for the opportunity to improve this figure, and we have addressed this with the other reviewers as well as there was an error with Figure 4 X-axis. It is not log-transformed, rather it was still on its scaled axis. We have rectified this error by making the X-axis the same as Figure 3 and correcting the figure caption. We added the identifications (a) and (b). Line 260 and 262 are referring to Figure 2 (a,b c, d).

5) Thank you for the questions. Yes, we did a thorough exploration of the data and identified likely thresholds for outliers, which are what we included. We would like to clarify that we allowed trips over 10km, rather we did not allow steps, i.e., the straight line connecting two consecutive GPS locations, to be longer than 10km, as these were unlikely to represent actual paths taken by the student.

Round 2

Reviewer 1 Report

The paper may now be accepted in its current form.