半数致死量LD_(50)的百分偏差和渐近方差

Percentage Bias and Variance of LD_50

毒性试验中的剂量与致死百分率平方根反正弦变换仍呈Ｓ形，且方差趋于同质， 半数致死量ＬＤ_(５０)可表示为 Ｌｏｇｉｓｔｉｃ模型参数的函数·因而，通过参数估计值的偏差和渐 过协方差矩阵，可推导出半数致死量ＬＤ_(５０)的百分偏差，渐近方差和ＬＤ_(５０)的置信区间．通 过 ２７个实例分析后建议，若 ＬＤ_(５０)的百分偏差超过 ５％，则 ＬＤ_(５０)

After arcsin-transformating the lethal percent, the error variance of the transformated lethal percent (TLP) was homogeneity and the curve of TLP on dose was S-shape, which is described by Logistic model . Therefore, median lethal dose (LD_50) is a function of the above parameter a and b, and the biases and covariance metrix of the estimated values of the above parameters could be estimated so that the errors of the estimated value of LD_50, i'e. percentage bias, variance and confidence intervals,- were obtained. Finally, the results from 27 examples showed that the error of the estimated value of LD_50 was large if the percentage bias of the estimated value of LD_50 exceeded 5%.