Design Patterns in Software Maintenance: An Experiment Replication at Freie Universität Berlin

An article published in 2001 reported a controlled experiment that compared maintenance of small programs using design patterns with maintenance of equivalent programs using simplified design solutions. A replication of that experiment was published in 2004. In 2010, a group of researchers from multiple countries picked this experiment as the subject of an attempt to perform a joint replication: Many groups performing an experiment using the same setup, each contributing a few data points to a larger overall data set. This article reports on one of those sub-replications. Only one of the results is statistically significant; it confirms the result of the original experiment stating that the simplified version of the GR program could be extended more quickly than the pattern version which used the Abstract Factory and Composite patterns. The article's main contributions, however, are (a) its description of the peculiarities of this particular subdataset and (b) its (implicit) suggestions for possible evaluation methods.