Inliers in a data set are observations which appear to be too small compared to the remaining observations in life testing with failure time distribution (FTD), F(x, 9), x > 0. To accommodate this situation, the article uses the model Mk(E), the exchangeable model introduced by Barnett and Lewis (1984) in connection with outliers, where it is assumed that (Xi, X2,..., Xn) are distributed independently so that A'of these observations are distributed as G(x, 0) and the remaining (n - k) are distributed as F(x, (0)). After quantifying the masking effect, the article provides an illustration assuming single inlier in case of exponential distribution for the Dixton type and Cocnran's test. (21 refs.)
展开▼
