交叉試驗設計(crossover design)於慢性疾病的臨床試驗已越來越常見，近年在R個序列與C個時間點的交叉試驗設計下於個數型資料的研究多假設資料源於卜瓦松分配(poisson distribution)進行統計推論，然而，在未知個數型資料真實的分配或相關性時，若以卜瓦松分配作為模型假設很可能得到不合適或不正確的結果。 本文利用強韌化的獨立卜瓦松實作概似函數，分析R個序列與C個時間點的交叉試驗設計下的個數型資料，以模擬研究與實例分析呈現強韌華德檢定統計量 (robust wald statistics)、強韌分數檢定統計量 (robust score statistics)、強韌概似比檢定統計量 (robust likelihood ratio statistics)，在未知資料真實分配為何的狀況下，仍可得到正確的統計推論。 關鍵字：交叉設計、強韌概似函數、相關性個數型資料 ;Crossover design has been often employed to study the effect of a treatment on chronic diseases, most research proposed parametric methods to analyze correlated count data under R-sequence and C-period crossover design based on Poisson distribution recently, however, the conclusions may collapse when we can’t specify the true underlying distributions. Hence, we provide a parametric method through the adjusted profile likelihood function to make Poisson likelihood robust. The corrected Poisson likelihood function can deliver legitimate likelihood inferences for parameters we are interested in. We demonstrate the simulation result that we can make statistical inferences correctly even though we use a misspecified model and use real data analysis to compare the robust Wald test statistic and robust score test statistic at the end. Keywords： Crossover design, Robust likelihood function, Correlated count data
