Abstract:
The objective of this study was to compare the computation of noncentral chi-square cumulative distribution, applied in sociology and engineering for example, used the noncentral chi-square in a radar scintillation study, using Patnaik method in the case of any degree of freedom (the first method) with three estimation methods, namely, the Ashour & Abdel-Samad method using any degree of freedom (the second method), the Ashour & Abdel-Samad method with odd degree of freedom (the third method) and the Ruben and Bock & Govindarajulu method in the case of odd degree of freedom (the fourth method). The average relative error of each methods was computed in order to compare its efficiency. The degrees of freedom using in this study were 1, 3, 5, 9, 15, 25, 35, 49 and 99, respectively. Noncentrality parameters (δ) and random variables (X) were obtained through uniformly distributed random number. From the results of the study, it can be concluded that the average relative error estimated from the second method is the least (0.0014), followed by the third (0.1496) and fourth one (31.8543), respectively. In case of high degree of freedom and low value of δ and X estimation, the results of the fourth method are far from exact value but approach exact value when using high value of δ and X. However, the results from each methods are rather close to one another. Therefore, the average relative time is considered as an additional contribution for this efficiency test and to control language in the calculation, the format of programming, and size of the object program. The results show that the fourth method is the fastest one. The followings are the second and third method, respectively. At present, the suitable decision will help magnificently in utilizing the limited resources efficiently. Therefore the second method seems to be the most suitable one for high precision, while the fourth method is the fastest one for the low degree of freedom calculation.