IWPSE 2015 - 2015 International Workshop on Principles of Software Evolution

The acceptance of empirical hypotheses should be supported by several and diverse types of replications. Replication studies make a critical evaluation of previous empirical studies for overlooked factors of the initial studies, for instance:
a) initial assumptions have evolved / changed / are not relevant anymore.
b) limited generalization, including: aspects that were previously ignored but that may affect the conclusions, conclusions that might only apply to certain types of applications, relations that have only been tested in certain types of applications but there are reasons to believe that it will not hold for other types of applications, etc.
c) examples of inconsistent or divergent results across studies about the same phenomena, including alternative hypotheses or differences in the experiment that could explain the divergence of results.
There are different types of replication depending on whether or not the experiment (model, variables and statistical method) and the data are the same. For instance,
a) Same experiment & Same data: aimed at evaluating the certainty of current knowledge (i.e., confirming or disputing previous results).
b) Different experiment & Same data: aimed at improving the model.
c) Same experiment & Different data: aimed at identifying limitations to the generality of the conclusions (or to problems with the data).
The first two types of replication are likely to be underreported due to the low contribution attributed to confirmatory results, and to their obliviousness with issues of the initial studies (either from the data or from missing variables). Nevertheless, they are crucial to understand the sensitivity of the conclusions to the variables analyzed.
We are also interested in meta-studies that analyze the replicability of empirical studies on software evolution, i.e., are all the details needed reported? at sufficient detail? to what extent original data can be reanalyzed? and to what extent the original experiment can be followed?