Page 144 -

130 CHAPTER 5 Surveys

5.13 SUMMARY

Surveys are a very powerful tool for collecting data from many individuals; how-
ever, there are only certain types of research questions in HCI for which surveys
are the most appropriate research method. For an appropriate survey method, there
must be a number of different steps that take place. To ensure validity and reliabil-
ity, survey questions must be pilot tested, to ensure that they are clear, unambigu-
ous, and unbiased. The overall survey design should make it easy for respondents
to understand and use the instrument, whether web-based, e-mailed, or on paper.
Appropriate sampling methods, even if they are nonprobabilistic, must be used to
ensure a representative response that can answer the research questions. Good intro-
ductions, establishing the credibility and importance of the survey, as well as pro-
viding ongoing reminders to respond, can increase the likelihood that there will be
a sufficient number of responses for the data to be considered valid. Other research
methods can also be useful in conjunction with surveys, such as focus groups, in-
terviews, or time diaries.

DISCUSSION QUESTIONS

1. Is a survey the same thing as a questionnaire? If not, how are they different?

2. What is the difference between the target population and the sampling frame?
3. Why are censuses done rarely? What is often used instead when population
estimates need to be made?

4. What is the defining characteristic of a probability sample?
5. What is a stratified random sample? How is it different from a traditional
random sample?

6. What is one of the major reasons that nonprobabilistic sampling is considered
appropriate in human-computer interaction research but not in other research
communities?
7. What is oversampling and why might it help improve validity of the research?
8. What is the difference between an open-ended and a closed-ended question?
9. Why might you want to use an existing survey instrument, when possible?

10. What is a double-barreled question and why is it not a good idea?
11. What is a contingent question and how might you deal with one in a survey
layout?

12. What are two methods for testing a survey tool?