Page 284 - Special Topic Session (STS) - Volume 2

STS493 Sofie d.B. et al.
and sensor data that are promising. In this paper, we will identify such
combinations in an official statistics context.
We see two approaches to employing sensor data: self-initiated and
existing sensor data. Under the first approach, sensor data do not yet exist
but are collected by responding businesses, persons or households. Under
the second approach, consent is asked to link existing sensor data that
businesses, persons or households have collected or are collecting. The two
approaches are not disjoint; in fact, they can be combined in a single data
collection. They can be further enriched using survey data. As a result, hybrid
data collections arise.
In the subsequent sections, we introduce criteria to evaluate the utility of
sensor data, briefly describe various types of sensors, illustrate ideas through
two examples and end with a discussion on future research.

2. Criteria for sensor data
Obviously, the mere possibility to collect sensor data is not enough reason
to also do so, as it requires a separate and new architecture and infrastructure
and respondents may not be willing to share the data. On the data collection
side, sensor data, especially when they are collected invitation by respondents
(i.e. primary data collection), demand for new data collection channels. These
channels demand for new and/or additional processing tools and skills, for
expansion of existing monitoring and analysis tools and for a redesign of
survey estimation methodology. Such changes are costly and time consuming.
On the respondent side, sensor data may still be burdensome and/or may be
privacy intrusive. Hence, a strong business case for mobile device sensor data
is needed and respondents need to benefit as well.
Sensor data are candidates to enrich or replace survey data when a survey
is relatively costly and/or inaccurate, when sensors are relatively cheap and/or
accurate, and when respondents are willing to share sensor data. We,
therefore, consider criteria from three perspectives: the survey quality-costs,
the sensor quality-costs and the respondent.
From the survey point of view, existing survey topics may be candidates for
enrichment or replacement with sensor data, when they satisfy at least one of
the following criteria:
 Burden: The survey topic(s) are burdensome for a respondent, in terms
of time, cognitive effort, or data retrieval;
 Centrality: The survey topic(s) are non-central to respondents, i.e. the
average respondent does not understand the question or does not
know the answer;
 Concepts: The survey topic(s) do not lend themselves to a
conceptualisation through a survey question-answer approach to
begin with;

273 | I S I W S C 2 0 1 9