Page 184 -

7.6 Research questions and hypotheses 171

3
Roughly speaking, there are four components of a case study design :
• questions;
• hypotheses or propositions;
• units of analysis; and
• a data analysis plan.

Research questions describe the goals of your study—what you are interested in
understanding.
Hypotheses or propositions are statements of what you expect to find. The unit
of analysis defines the granularity of your study—what exactly you are focusing on.
Are you studying an organization, a group of people, an individual, or individual
activities? These questions will guide your data collection. The final component—a
data analysis plan—is described in Section 7.8.
Just as in other forms of research, your research questions and hypotheses guide
your efforts. You may be interested in understanding how users accomplish certain
goals or tools, how the introduction of a new tool changes the workflows and patterns
in an organization, or what a team needs from a new collaboration tool. Even if your
case study is exploratory or descriptive, you should try to make your research ques-
tions and propositions explicit.
Taken together, your research questions and hypotheses form a preliminary model
that will guide your development of the case study. By mapping out your interests
and the range of concerns that you are trying to address, you will gain greater under-
standing of the criteria that you will use to choose your cases, the data sources that
you might need to include, and how you will conduct your analysis. The approach
of ignoring theory in favor of simply collecting data indiscriminately can be a recipe
for failure (Yin, 2014).
In Sara's case study, the researchers were interested in understanding how a blind
person might use a variety of assistive technologies to accomplish tasks and to recover
from task failures using workarounds. These questions led to several propositions. The
investigators expected to see common types of failures and workaround strategies.
They also expected that the choice of implementing features in hardware or software
might influence user interactions, including failures and responses to those failures.
A different set of research questions might have led the researchers to a very
different case study. If, for example, a preliminary study had led them to believe
that education or socioeconomic status might play an important role in determining
how blind people use technology, they might have chosen a multiple-case design,
including participants with backgrounds that differed in these relevant respects. They
might also have asked a broader range of questions about background and included
consideration about other aspects of their participants' lives.

3 This list is based on Robert K. Yin's list of five components. His list divides the “data analysis plan”
into two components: the logic linking the data to the propositions and the criteria for interpreting the
findings (Yin, 2014).