A method for measuring programmer debugging performance from key strokes
Y. Takada, Koji Torii · Systems and Computers in Japan · 1995
Abstract In software development management, a method is necessary that is capable of measuring programmer performance objectively, and easily. This paper presents a method to measure programmer debugging performance from key strokes. Debugging performance is defined as the efficiency of programmer activities from discovery of software failures to correction of their causes. In this method, programmer activity is first categorized into several classes by monitoring key strokes. Programmer property parameters are then extracted by applying model‐based analysis on monitored programmer activity sequences. The programmer model defined here is based on the assumption that a programmer activity sequence is a Markov process. Finally, several programmer performance values are computed from the extracted parameters. The major value is D, which is the estimated length of time required for debugging one fault. Experiments show that the method is effective enough in practical software development. One of these findings is that D shows strong correlation with a value that represents programmer productivity (the length of program development time for common specifications).