Using peak analysis for identifying lagged effects between software metrics

Measures extracted from software repositories tend to be collected on a regular basis (often daily), forming time series of data. In this context, it is normal to assume that some of the measures collected are having an effect on other measures, potentially with some delay (or “lag”) in the effect. Such delay in the effect may even vary over time, making the identification of the effect difficult. In this paper, we present our initial ideas on a simple analysis method for such situation, in which peaks from the two time series in question are analyzed, and similar ones are