วิธีการประมาณความน่าจะเป็นที่จะเสียชีวิตสำหรับข้อมูลประกันชีวิต ที่ไม่สมบูรณ์ / สมบัติ กุลวุฒิ = Estimation methods of mortality probability for incomplete life insurance data / Sombat Kullawoot
The objective of this study is to compare estimation methods of mortality probability for incomplete life insurance data. The estimation methods under consideration in this study are Classical Estimation Method, Maximum Likelihood Estimation Method and Bayes Estimation Method with Exponential Distribution as a Prior. Each method estimates the probability that a person, whose age is x, will die within one year (qx). The ages studied are between 25 and 65 years old, inclusively. The sample sizes (m) are 30, 50, 70, 100, 500, 700 and 1000. The distributions of future lifetime for this study are Weibull, and Gompertz. To estimate qx for each sample sample size and each distribution, the experimentations are repeated 500 times using the Monte Carlo simulation method, and their the absolute percentage errors (APE) are evaluated. The results of this study are as follows: For each distribution and for all sample sizes Bayes Estimation Method has the lowest APE and the Classical Estimation Method has the highest APE. Comparing only two methods, Maximum Likelihood Estimation and Classical Estimation, The APE’s are not much different in all conditions. When the sample size increases, the APE’s of all three methods decrease, and when the sample size is very large (m = 700 and 1000), the differences in magnitude of APE’s of all methods tend to become small.