| 摘要 |
PURPOSE:To realize the titled check system with an error distribution produced by the sequence taken into consideration by replacing a software reliability growth curve with a standardized double exponent distribution and utilizing the property where each probability density distribution of the sequence statistic value can be approximated to a normal distribution. CONSTITUTION:An optimum model is selected out of various models of software reliability growth curves based on Akaike's Information Criterion. This selected model is applied to the observation data for estimation of the parameter of a model formula. An optimum model formula is replaced with a standardized double exponent distribution by means of the parameter estimation value. Then the observation data is converted into the data subject to a standard normal distribution via the average and dispersion of the probability density distribution of the sequence statistic value subject to said exponent distribution. It is checked whether the converted data train is subject to the standard normal distribution or not by means of a smooth test of Neyman or the check of Shapiro-Wilk. When this check is successful, the limit of reliability is obtained for the bag estimated value at an optional time point of observation.
|
