การเปรียบเทียบวิธีการประมาณแบบช่วงสำหรับพารามิเตอร์ของการแจกแจงพหุนาม / บัณฑิต ฉัตรตะวัน = A comparison on methods of interval estimation for parameter of multinomial distribution / Bandit Chattawan
The objective of this research was to compare the interval estimation for the parameters of multinomial distribution. The estimation uses method of Normal, Quesenbery and Hurst's and F with sample of size 50, 100, 200, 500 and 1000 from multinomial distributions of groups 4, 6, 8 and 10. For each group we divided into three cases; equality, bias, and symmetry. All of which were considered at 90%, 95% and 99% confidence levels.The data in this research were obtained from MATLAB program by using Monte Carlo technique and this process was repeated 2000 times. The conclusion of this study are as follow: 1. The Normal method performs the lowest average confidence levels and the length in term of average confidence region length. The average confidence levels are not lower than the given confidence levels when the sample size is bigger than 500 for 4 groups and bigger than 1000 for 6 groups from 4, 6, 8 and 10 groups of proportions. 2. The Quesenberry an Hurst method performs the highest average confidence levels and the length in term of average confidence region length. The average confidence levels are not lower than the given confidence levels. 3. The F method performs the average confidence levels higher than the normal method but its length in term of average confidence region length is near the normal method's. The average confidence levels are not lower than the given confidence levels