Page 129 -

5.5 Nonprobabilistic sampling 115

that, in some cases, researchers may have a goal to get representative data from mul-
tiple countries to do a multinational comparison. A great example of this is Harbach
et al.'s (2016) recent study examining smartphone locking in eight different countries,
with over 8000 survey responses. In such cases, it may be necessary to collect detailed
demographic and cultural data related to the country, and researchers are also encour-
aged to consult a guide on doing cross-cultural HCI research (e.g., Aykin, 2005).

5.5.2 OVERSAMPLING
When there is not a well-defined list of users and strict random sampling is not pos-
sible, then the number of responses becomes increasingly important. For instance, in
a nonprobabilistic sample, 20 survey responses may not be sufficient. Even with de-
mographic data present, there may just be too many biases present, relating to which
users have responded. However, when the survey response reaches a certain number
that is considered large in proportion to the estimated or perceived population size,
this can help establish some informal validity. This is known as oversampling. While
not all researchers agree that oversampling increases validity (Couper, 2000), simply
having a large response can reduce the likelihood of excluding any segment of the
population (Andrews et al., 2003). However, the key is that the response must be
large in the context of the population of interest. For instance, 500 survey responses
would be a large number if the estimated total population of interest is around 5000
individuals. However, 500 survey responses would not considered large if the popu-
lation of interest is a country, such as Australia or France. One researcher suggests
that 30 responses should be considered a baseline minimum number of responses
for any type of survey research (Sue and Ritter, 2007). Fogg et al. used both de-
mographic data and oversampling to learn more about web credibility in 2001 (see
Demographic Data and Oversampling sidebar).

DEMOGRAPHIC DATA AND OVERSAMPLING
Fogg et al. (2001) wanted to learn more about how different elements of design
on a website impact on the user's perception of credibility. To do this, they
recruited survey responses through charitable groups in the United States and
a news media organization in Finland. They received 1441 survey responses in
1 week. After discarding a number of responses due to inadequate information
provided or responses that placed the respondent outside of the population
frame, 1410 survey responses were considered valid.
The survey collected information on age, gender, country, education level,
income, years on the Internet, average number of hours spent online per week,
and average number of purchases online. The demographic information helped
to confirm that the responses to the survey were, indeed, representative of the
diversity of web users. The high number of responses helped to improve the
validity of the study.