Q:What is a research design?

A:A framework for every stage of the collection and analysis of data

A research design provides a framework for the collection and analysis of

data". The choice of methods to be used is, indeed, very important, as is
an understanding of your fundamental research philosophy. But a
research design will highlight these choices and other decisions about
which elements are considered to be more important than others, as well
as your hypotheses about causality and predictability. Consider it as a
blueprint for the research you propose to conduct. This chapter looks at
five different research designs from which you could choose.

 If a study is "reliable", this means that:

 The measures devised for concepts are stable on different occasions
 The essential question about research is its reliability. It is often the
case that concepts in the social sciences can be construed differently
in different social contexts, so the promise of repeatability makes
readers feel the results can be relied on more. But what is even more
important is that there should be not much variation (or none at all) in
responses to the same instruments by the same type of respondent.
Bryman gives the example of wild fluctuations in IQ test scores as an
indicator of low reliability of the test itself. When reviewing literature
or consulting secondary sources, we are certainly influenced by the
reputation, or simply good standing in the academic community, of
the researcher. This does not imply uncritical acceptance of their
findings, however.

 "Internal validity" refers to:
 Whether or not there is really a causal relationship between two
variables
 "Validity" has a special meaning in research, usually indicating the
truth of something, its authenticity. Many of our research activities
can be seen as valid steps towards producing a dissertation, for
example, but our conclusions will not be worthwhile unless our
research was valid. If a measure proves unreliable (see question 2), it
lacks "measurement validity" but "internal validity" is lost when the
"internal" relationship between variables is lost, or ambiguous, or
confused. Typically, we argue that "a" causes "b", but if "b" can
 actually influence the value of "a", then the causal relationship
suggested doesn't really exist.
 Lincoln & Guba (1985) propose that an alternative criterion for
evaluating qualitative research would be:
 Trustworthiness
 Most tests of reliability and validity are applicable to quantitative
data rather than to qualitative. Lincoln and Guba (1985) propose
"trustworthiness" as an example of a criterion that could determine
how good the qualitative research might have been. This criterion
may be subdivided into dimensions of credibility, transferability,
dependability and confirmability (which Bryman examines in detail
in chapter 15), to act as counterparts for reliability and validity in
quantitative research. It is the view of many that whereas running a
focus group, for example, may be 'messier' than conducting a survey,
messiness should not be a goal of the research!

 Naturalism has been defined as:
 Key concept 3.4 explains that "naturalism" is an unusual expression
which has many meanings, some contradictory! All of the definitions
shown in this question are correct, although "a" is positivist as
opposed to the interpretivism suggested by "b" and "c". However,
research methodologies like ethnography, or observation, or
unstructured qualitative interviews try to come close to the natural
context of the data, while being relatively non-intrusive.
 In an experimental design, the dependent variable is:
 The one that is not manipulated and in which any changes are
observed
 When conducting an experiment, it is essential to manipulate one
variable, (conventionally called "independent") so that changes in
another (the dependent variable) can be identified as indicating a
causal relationship. There is nothing ambiguous about this process in
the slightest, nor do personal values intrude. Recalling that many
"independent variables" cannot be manipulated in an actual social
context, experimentation may be the only way of getting close to an
identification of a causal relationship between variables
 What is a cross-sectional design?
 The collection of data from more than one case at one moment in
time