การเปรียบเทียบอำนาจของการทดสอบสำหรับสัมประสิทธิ์ในตัวแบบอัตถดถอยอันดับ 1 แบบมีแนวโน้ม / สุทธาทิพย์ มาระเนตร์ = A comparison of power of the tests for a coefficient in first-order autoregressive model with trend / Sutthati Maranate
The objective of this study is to investigate the probability of type - I error and power of the test for a coefficient in first order autoregressive model with trend. The test statistics are t-test, Bootstrap t-test and Dickey - Fuller test under conditions of severity of autocorrelation coefficient of dependent variable (yt), sample sizes and distributions of random error (Ut). The data for this experiment were generated through the Monte Carlo simulation technique by personal computer. It was used to calculate the probability type - I error and power of the test. The experiment was repeat 1,000 times under each condition at one percent, five percent and ten percent significant level. Results of the study are as follow : - 1) Probability of type - I error : t-test could control the probability of type - I error except sample sizes are 15 and 20 for all significant levels. Bootstrap t-test could control the probability of type - I error for all simulation conditions in this study. Dickey - Fuller test could control the probability of type - I error except sample sizes is 15 at 1% significant level. 2) Power of the test : Bootstrap t-test and Dickey and Fuller test had a nearly high power for almost simulation conditions. T-test had the lowest power for all simulation conditions in this study.
