A model of noisy software engineering data

Software development data is highly variable, which often results in underlying trends being hidden. In order to address this problem, a method of data analysis, adapted from the financial community, is presented that permits the shape of the curve of some activity to be reduced to a few line segments, called the characteristic curve. This process is used on sample data from the NASA/GSFC Software Engineering Laboratory and has been shown to be a reasonable method to understand the underlying process being plotted