藉由模擬研究與實例分析，比較強韌華德檢定統計量、強韌分數檢定統計 量、強韌概似比檢定統計量、Seen (1994) 提出的 CROS 檢定統計量和 Putt and Chinchilli (2004) 提出的魏克生符號檢定 (Wilcoxon signed rank)。不論實作模型與資料間的假設是否相同，透過強韌概似函數方法可以得出正確的統計推論。;In clinical trials, the experimental technique commonly used by experimenters is a crossover design. In this paper, independent normal model and independent gamma model are used as working model, and the robust likelihood function method is used to robust working model. Further obtain the robust Wald test statistics , the robust score test statistics , the robust likelihood ratio test statistics , and confidence intervals for individual tests are also available. By using simulation research and real data analysis, we can compare the robust Wald test statistic, robust score test statistic, robust likelihood ratio test statistic, CROS test statistic proposed by Seen (1994), and Putt and Chinchilli (2004) proposed Wilcoxon signed ranked. Regardless of whether the assumption between the working model and the data are the same, robust likelihood function method can obtain correctly statistical inferences.
