The Importance of Rating Level Selection for Input Variables to Determine Accurate Estimations in Parametric Mathematical Models
Juan José Cuadrado-Gallego, Óscar Marbán, Antonio de Amescua Seco, Marı́a-Isabel Sánchez-Segura · The Journal of Cost Analysis & Management · 2004
The use of parametric models is one of the main cost estimation methods in software engineering. It is thought that precision in these models depends on three fundamental factors: the mathematical equations applied; the calibration of the numerical values, which may take the input variables used by the model; and a correct rating level selection for each input variable for each project. This study shows that for accurate estimations, the third factor, the correct rating level for input variables of a model, may have as much or more importance than the other two. This is a study of the numerical relation of the errors that may be introduced in the estimations through each of the three factors.