Variance estimate for up and down test

Traditional up-and -down statistical methods for sensitivity test, are appropriate for estimating the mean but not the variance of sensitivity variables. The estimating value of variance by traditional methods is generally lower than the actual one and the number of samples required by these methods is quite large. In this paper, a estimate formula for the variance of sensitivity variables is studied in detail. This formula fully considers the effect of test conditions on the estimating accuracy, resulting in higher precision. In order to give an objective demonstration, we compared the present method with Dixon method by Monte Carlo analysis and showed relevant results. The estimating values by the present method are much more accurate than those by Dixon method under different conditions. Even with small sample number, the present method still could obtain a good estimate value. The proposed method not only solves the problem in traditional methods but also reduced the number of specimens, resulting in great saving. Finally, an application for AC9 aluminum alloy is given.