Type II censored samples have been extensively collected from many fields including engineering and biomedical sciences. These samples frequently contain extreme values and the predictive inference from such type of samples may produce invalid results. In this paper, we compare predictive variability with respect to the location parameter mean as well as median. A simulation study and a real life study are considered to determine the predictive results. It is obtained that the predictive estimates with the location parameter median give more precise results compared with the location parameter mean when type II censored sample contains extreme values.

censored sample, half-normal model, statistical inference, predictive inference.