Missing Data and Other Measurement Quality Issues in Mobile Geolocation Sensor Data

Abstract

As smartphones become increasingly prevalent, social scientists are recognizing the ubiquitous data generated by the sensors built into these devices as an innovative data source. Passively collected data from sensors that measure geolocation or movement provide an unobtrusive way to observe participants in everyday situations and are free from reactivity biases. Information on day-to-day geolocation could provide valuable insights into human behavior that cannot be collected via surveys. However, little is known about the quality of the resulting data. Using data from a 2018 German population-based probability app study, this article focuses on the measurement quality of geolocation sensor data, with a strong focus on missing measurements. Geolocation sensor data are an example of an available data type that is of interest to social science research. Our findings can be applied to the wider subject of sensor data. In our article, we demonstrate (1) that sensor data are far from error-free. Instead, device-related error sources, such as the manufacturer and operating system settings, design decisions of the research app, third-party apps, and the participant, can interfere with the measurement. To disentangle the different influences, we (2) apply a multistage error model to analyze and control the error sources in the specific missingness process of geolocation data. We (3) raise awareness of error sources in geolocation measurement, such as the use of GPS falsifier apps, or device sharing among participants. By identifying the different error sources and analyzing their determinants, we recommend (4) identification strategies for future research.

Using a Mobile App When Surveying Highly Mobile Populations: Panel Attrition, Consent, and Interviewer Effects in a Survey of Refugees

“…Additionally to using mobile phone applications for interviews, smartphones are also used to implement geo-tracking of individuals, as the majority of current devices are equipped with GPS sensing (for a discussion see Bähr et al, 2020). Besides gathering information on individual movement, researchers are increasingly able to use geo-spatial data to learn about, for instance, the respondents' place of work, or their leisure activities (Colak et al, 2016).…”

Augmenting Surveys With Data From Sensors and Apps: Opportunities and Challenges

“…Apps can aggregate and process this information, store it, and transfer the data to a database outside the phone. For example, studies by Bähr et al (2020) and English et al (2020) in this special issue use apps that administered questions to respondents at different moments throughout the day.…”

“…Sensor data can be collected via apps that are installed on the participants' device or via Internet browsers (e.g., in a mobile web survey). The app-based approach allows for a relatively continuous data collection during a comparatively long time period (e.g., Bähr et al, 2020). For instance, for GPS data, researchers can track geographical motion profiles of participants to infer their travel behavior.…”

“…For example, the project described in English et al (2020) in this special issue used environmental sensors that collected information about temperature, pressure, humidity, and air pollutants, and some sensors collected information about sound, vibration, and imagery. Bähr et al's (2020) app described in this special issue collected data on geolocation, physical activity, device use, communication patterns, and ethnic composition of social networks of the participants based on name classifiers.…”

“…The task of downloading an app for data collection might involve further potential self-selection effects as it requires additional steps from participants. Studies recruiting their participants from population-based panels report download rates of under 20% (Bähr et al, 2020;Jäckle et al, 2019;Kreuter et al, 2020). For a cross-sectional study using the Dutch CBS Travel App, participants were recruited using the population register, and 35% of the invited sample members downloaded the app (McCool et al, forthcoming).…”