ESRA 2019 Programme at a Glance


Survey and Spatial (gps) Data. Understanding Human Behavior in a Spatial Context

Session Organiser Mr Peter Lugtig (Utrecht University)
TimeTuesday 16th July, 16:00 - 17:00
Room D21

Apps on smartphones or other wearable devices enable us to follow people over time, and track what they do, where they go, at what time. The time-location data that result from such devices are now starting to replace survey data (e.g. for a travel diaries), or provide a rich context to existing survey data (e.g. in time use research).
More broadly, they allow us to understand human behavior in a better way: we can use techniques from GIS and spatial registers to enrich time-location data. A specific location can for example be at a train station, a particular shop or place for leisure. We can ask respondents about experiences of having been at a particular location. As another example, we can combine attitudinal survey data with behavioral location data to for example understand transportation choice.

Combing geodata with survey-data, both during data collection and in the analysis, offers lots of potential for future social science research. There are however still lots of practical and methodological problems in using and combining survey and location data. This session focuses on emerging methodologies in dealing handling location data collected through wearables and survey data. We invite proposals for papers focused on, but not limited to:

- Building apps or hardware to measure time-location data for people
- Methods to effectively record location-data in these apps
- Using location-data to ask specific survey questions during data collections.
- Enriching location-data using registers or GIS
- Combining survey and location data during data analysis.
- Studying measurement error properties of location-data.
- Dealing with missing episodes of location data.

Keywords: gps, geodata, locations, survey, time use, travel

Automatic Trip and Transportation Mode Detection Using a Smartphone App and Machine-Learning. A Validation Study.

Mr Laurent Smeets (Utrecht University)
Dr Peter Lugtig (Utrecht University) - Presenting Author
Professor Barry Schouten (Statistics Netherlands)
Mrs Danielle McCool (Utrecht University)

Smartphone apps hold the promise of improving the way we collect data. In November 2018, Statistics Netherlands and Utrecht University conducted a large experiment among 1900 sample members with a smartphone app that automatically records stops and trips over a one-week period. The goal of the app is to see whether we can improve the quality of official statistics related to travel behaviour. Historically, travel statistics have mostly been collected with paper diaries, which we know suffer from recall error, and underreporting of short trips.

The app passively collects data on location using a combination of sensors (GPS, WiFi and cellular sensors), and presents respondents with a timeline per day of stops and trips made per day. The underlying location-measurements allow us to calculate how fast respondents have traveled and with what route, enabling us to classify whether respondents traveled for example on foot, by bike or car on a particular trip.

The classification algorithms used in transportation mode detection may in some cases lead to errors (e.g. when a car is stuck in traffic). Also, some modes of transport are generally hard to distinguish (e.g. bus/tram/car), limiting to some degree the power of smartphone apps to truly deliver better travel behaviour statistics.

As part of our field experiment, we re-interviewed respondents who completed a paper/web diary on travel behavior earlier in 2018. Furthermore, within the smartphone app, we asked respondents to voluntarily indicate for every trip which modes of transportation were used. Although these survey data also contain errors, they do enable us to train and test transport classification algorithms. This presentation will show whether passively collecting data indeed produce better statistics. In particular, we will focus on showing how we designed the app, and trained and tested a transportation mode algorithm to deliver more timely, accurate and richer travel statistics.


GPS-Paradata in Computer-Assisted Personal Interviews: Additional Opportunities for Monitoring Fieldwork Interviewers

Mr Daniil Lebedev (National Research University ) - Presenting Author

Download presentation

Wide proliferation of computer-assisted methods of data collection provide opportunities associated with paradata employment in order to get more insight on what and how is going on during fieldwork process. One of these possibilities is geo positioning system paradata (GPS-paradata). Data on the place of the interview or GPS-paradata is one the simplest and intuitively understandable ways of paradata utilization in case of fieldwork monitoring (interviewers control) which is though connected with many methodological challenges.
This paper is concerned with examination of the possibilities of GPS-paradata employment in the context of interviewers’ control with particular emphasis on important methodological problems which emerge in case of its employment.
We present here the examination of the quality of received GPS-paradata and results of fieldwork monitoring employed with GPS-paradata utilization. Two methods were used - analysis of discrete GPS-points to compare the place of interviewer in the start and in the end of the interview and analysis of GPS-points' sequence (interviewers' paths) to define the patterns of interviewers' routes. Future possible implications of such analysis are presented with a focus on combining such techniques with existing methods and taking into account the data quality.