Evaluating the relation between changeability decay and the characteristics of clones and methods

In this paper we propose a methodology to evaluate if there is a relation between two code characteristics. The methodology is based on relative risk, an epidemiology formula used to analyze the effect of toxic agents in developing diseases. We present a metaphor in which the disease is changeability decay, measured at method level, and the toxic agent is a source code characteristic considered harmful. However, the formula assesses the strength of the relation between any toxic agent and any disease. We apply the methodology to explore cloning as a toxic agent that increases the risk of changeability decay. Cloning is a good agent to analyze given that although there is some evidence of maintainability issues caused by clones, we do not know which clones are harmful, or to what extent. We compare cloning with other possible dasiatoxic agentspsila, like having high complexity or having high fan-in. We also use the technique to evaluate which clone characteristics (like clone size) may indicate harmful clones, by testing such characteristics as toxic agents. We found that cloning is one of the method characteristics that affects the least changeability decay, and that none of the clone characteristics analyzed are related with changeability decay.