CFE 2019: Start Registration
View Submission - CMStatistics
B1777
Title: Lower quantile estimation for artificially censored Weibull samples Authors:  Jarod Smith - University of Pretoria (South Africa) [presenting]
JT Ferreira - University of Pretoria (South Africa)
Andriette Bekker - University of Pretoria (South Africa)
Abstract: Quantile estimation is a vital aspect of statistical analyses in a variety of fields. For example, lower quantile estimation is crucial to ensure the safety and reliability of wood-built structures. An intuitive approach would be to consider models that fit the tail of the sample instead of the entire range. Quantiles of interest can be estimated by artificially censoring observations beyond a chosen threshold. The choice of threshold is crucial to ensure efficient and unbiased quantile estimates, and usually the 10th empirical percentile is chosen as the threshold. A bootstrap approach has been previously proposed in order to obtain a better threshold for the censored MLE, however, this approach is computationally expensive. A new threshold selection technique is proposed that makes use of a standardised-weighted adjusted truncated Kolmogorov-Smirnov test (SWAKS-MLE). The SWAKS-MLE outperforms in the bootstrap threshold censored Weibull MLE method, in addition to being vastly less computationally intensive.