2.2. Selective daily mobility bias in GPS studies
The description of behavioral contexts (e.g., of the places
actually used to exercise) can help plan the provision of healthenhancing
services (Duncan and Mummery, 2007; Lachowycz
et al., 2012; Quigg et al., 2010). Moreover, this descriptive
information is useful to generate hypotheses on environmental
resources that support behavior; these causal hypotheses then
need to be formally tested through appropriate designs. Our focus
here is on such causal inferences on environmental effects (‘‘does
the presence of X in the environment influence the behavior of
people who live, work, or spend time nearby?’’).
In the reviewed studies, exposures to the built and food
environments were determined around valid GPS points, including
places where individuals specifically go to practice the
behavior investigated (e.g., to buy or eat specific foods or to
practice physical activity). With such measures, participants with
a particular taste for energy-dense food that leads them to eat
several days a week in fast-food restaurants would be classified as
often ‘‘exposed’’ to a significant density of fast-food restaurants.
Similarly, people with advanced knowledge on the benefits of
physical activity, positive attitudes toward exercise, and a suffi-
cient self-efficacy to convert intentions into regular physical
activity more frequently visit sport or recreational facilities and
would therefore appear as more ‘‘exposed’’ to physical activity
opportunities. Such intrapersonal factors that encourage specific
study participants to regularly use the corresponding environmental
opportunities contribute to generate a relationship
between environmental factors and health behavior that could
be spuriously interpreted as a causal effect of the environment
(either in contemporaneous momentary or individual-level analyses).
Even GPS studies with refined research questions examining,
e.g., which types of greenspaces visited or which
characteristics of parks visited are associated with higher physical
activity levels do not allow the identification of causal environmental
effects because people select the type of parks they go to
according to their intended use of these facilities.