Distribution-free approach to the sum of correlated variates
- Bangkok : Asian Institute of Technology, 1982
- vi, 38 p.
- Thesis ; no. WA-83-14 .
- Asian Institute of Technology. Thesis ; no. WA-83-14 .
A thesis submitted in partial fulfillment of the requirement for the degree of Master of Engineering, School of Engineering and Technology
Thesis (M.Eng.) - Asian Institute of Technology, 1982
A method to determine the three statistics, the mean, variance and the skew coefficient, of the sum of correlated variates that are used to fit the data to three-parameter density functions was developed. A primary task that is integral to this study is the determination of the transformation matrix (matrix that transform correlated variables to uncorrelated variables) that would be reliable when being used in the computation of the moments, especially the skew coefficients. To determine the appropriate transformation matrix the comparisons between the results using the proposed approach and the results obtained via exact gamma solutions as well as Monte Carlo experiments were conducted. The three- parameter log-normal distribution (LN3) was used to model the distribution of the random variables generated and AR(l) and ARMA(l ,l ) processes were used to resemble the correlation schemes of the variables in the sum in the Monte Carlo experiments. The moments computed through this approach were satisfactory. The transformation matrix PD1/2 PT was selected to be the most appropriate matrix especially in the computation of the third moment. The other transformation matrices were also found to be effective in the computation of the second moment.