Replication of Software Engineering Experiments - Empirical Software Engineering and Verification - page 84

the internal replication is built into the original study design. So the data, part
of which are used for the replication, are gathered simultaneously by the same
investigator using a common set of research operations. One finds variations
in the procedures for selecting the samples. Two of these procedures are: 1)
drawing two or more independent samples, and 2) taking a single sample and
later dividing it into subsamples for purposes of analysis and comparison. The
internal replication provides an additional data supply which acts to cross-check
the reliability of the observed relationships. Thus it is methodologically superior
to the single study where the hypothesis is tested only once by one body of data.
Independent Replication. Independent replication is the basic procedure for ver-
ifying an empirical generalization. It does this by introducing significant modifi-
cations into the original research design in order to answer questions about the
empirical generalization that go beyond those of reliability and confirmation. The
essential modifications include independent samples drawn from related or dif-
ferent universes by different investigators. These replications differ in design and
purpose. They can, however, be broadly categorized into three problem areas.
First, is the empirical generalization valid? Second, does further investigation
extend it to other social situations or subgroups outside the scope of the origi-
nal study? Or, third, is the empirical generalization limited by the conditions of
particular social situations or specific subgroups?
Theoretical Replication. It involves the inductive process of examining the fea-
sibility of fitting empirical findings into a general theoretical framework. These
replications seek to verify theoretical generalizations. In these replications, em-
pirical variables, which have concrete anchoring points are abstracted and con-
ceptualized to a higher theoretical plane, it is necessary to sample a variety of
groups using different indicators of the same concepts.
A.11 Leone and Schultz [73]
Experimental Replication. The same experiment is conduced more than once,
although there can be (especially with social systems) no perfect replications. It
involves the same method and the same situation.
Nonexperimental Replication. The same method is applied to different situations.
Corroboration. It involves different method and same situation, or different
method and different situation.
A.12 Lindsay and Ehrenberg [68]
Close Replication. This replication attempts to keep almost all the know conditions
of the study much the same or at least very similar (for example, the population or
populations in question, the sampling procedure, the measuring techniques, the
background conditions, and the methods of analysis). A close replication is par-
ticularly suitable early in a program of research to establish quickly and relatively
easily and cheaply whether a new result can be repeated at all.