兩藥品通過生物對等性檢定 (bioequivalence test) 後可稱之具生物對等性，一般假設藥物動力學 (pharmacokinetic) 資料服從對數常態分配 (log-normal distribution)，經過對數轉換後，以二元常態分配 (bivariate normal distribution) 為模型作檢定。但二元常態分配模型參數較多，計算繁雜，因此本文提出以二元負二項分配 (bivariate negative binomial distribution) 為模型。相對於二元常態分配，負二項分配參數較少且容易計算。將此模型適當修正後可得一具強韌性的概似函數，在資料分配不知的情形下，可方便的分析成對的資料 (paired data)，除可正確的估計參數外亦可得到正確的推論。;Only when the two drugs pass the bioequivalence test, can we claim that two drugs are bioequivalent. Usually, the distribution of the pharmacokinetic data is assumed to be log-normal and inference is made under normality with logarithmically transformed data. The number of parameters in bivariate normal model makes it less convenient to make inference about bioequivalence. We propose using the bivariate negative binomial model to test for bioequivalence. We can convert the bivariate negative binomial likelihood to become robust to accommodate general pharmacokinetic data whose distribution might be less understood.
