Page 143 -

5.12 Data analysis 129

4. Send a replacement survey to nonrespondents 2–4 weeks after the original one
was sent.
5. Make a final contact using a different mode. If the original survey was sent
using postal mail, then maybe a phone call or e-mail should be used. If the
survey was electronic, maybe a postal letter or phone call should be used. The
idea is to have a different delivery method for the final contact that gets the
attention of the respondent.
Depending on how the researchers have access to the potential respondents,
different methods of postal mail, e-mail, phone calls, or even instant messaging,
may be interchanged. So, for instance, in an electronic survey (web-based or e-mail),
multiple reminders are important, and each time, the researchers should give the
survey instrument (for a web-based survey).
A common question, mentioned earlier in the chapter, is the question “How many
survey responses are enough?” This is not easy to answer, as it has to do with a num-
ber of different issues: What is the goal of the survey? What type of survey? What
sampling method has been used? What level of confidence and margin of error is
considered acceptable? See Section 5.4.2 earlier in this chapter, where these ques-
tions are discussed.

5.12 DATA ANALYSIS
There are several ways to analyze survey data. The analysis chosen will depend, in
large part, on:
• whether it was a probabilistic or nonprobabilistic survey;
• how many responses were received; and
• whether a majority of questions were open-ended or closed-ended questions.
Generally, the quantitative and qualitative data is separated for analysis. The data
is “cleaned,” meaning that the researchers look through and make sure that each
survey response is valid, and that none of the responses are either repeats (where the
same person submitted more than one response), incomplete (where most questions
were not answered), or invalid (due to a respondent not meeting the qualifications).
The quantitative data is ready to analyze, whereas the qualitative data must first be
coded (see Chapter 11 for more information on content analysis).
Often, the goal of quantitative data analysis is simply to have a set of “descrip-
tive statistics” that simply describe the data collected in a manageable way (Babbie,
1990). No one but the researchers will read through every survey response so the
descriptive statistics are simply a short, high-level summary of the data. Most often,
descriptive statistics involve percentages, ratios, or matrices. Inferential statistics in-
volve a higher level of understanding of the data, by understanding the relationships
between variables and how they impact each other. For more information on statisti-
cal analysis, read Chapter 4.