Replication of Software Engineering Experiments - Empirical Software Engineering and Verification - page 80

in which every nuance of the experimental setting is precisely reproduced; as
such, the cause-effect relationship is finite. The ability to conduct a Type 0
replication is limited to experimenters in only some of the natural sciences. As
others have stated, it is an impossibility to conduct a Type 0 replication in a
social science context because uncontrolled extraneous factors have the potential
to interact with the various components in an experimental setting. For example,
human subjects cannot be precisely duplicated. A social scientist is limited only
to matching subjects as closely as possible.

Type I (Duplication). A type I replication is a faithful duplication of a prior
study and, as such, is considered the “purest” form of replication research in the
social sciences. It should be mentioned at this point that a Type I replication is
the one most closely associated with the term “replication” in the minds of most
researches. More over, this is also the type of replication research most criticized
for not being creative. This is somewhat ironic, given the apparent receptivity
of reviewers to cross-cultural research that, in many cases, is usually the study
of the generalizability of findings from a single country or culture to others and,
thus, is simply a Type I replication.

Type II (Similar). A type II replication is a close replication is a close replication
of a prior study, and a Type III replication is a deliberate modification of a prior
study. Type II replications are the most common form of replication research
in marketing settings and are useful in testing phenomena in multiple contexts.
If effects are shown in a variety of testing contexts, the case for the findings is
strengthened. This has been called the process of triangulation.

Type III (Modification). This replication is a deliberate modification of a prior
study. In a Type III replication, the threat of extraneous factors inherent to the
nature of human subjects, unless explicitly accounted for in theory testing, is
not a factor of concern with regard to replicability.

A.3 Evanschitzky and Armstrong [64]

Real Replications. This replication is a duplication of a previously published
empirical study that is concerned with assessing whether similar findings can
be obtained upon repeating the study. This definition covers what are variously
referred to as “exact”, “straight” or “direct” replications. Such works duplicate
as closely as possible the research design used in the original study by employ-
ing the same variable definitions, settings, measurement instruments, analytical
techniques, and so on.

Model Comparisons. This replication is an application of a previously published
statistical analysis that is concerned with assessing whether a superior goodness-
of-fit can be obtained, comparing the original statistical model with at least one
other statistical model.