The main objective of JPS and ranked set sampling (RSS) designs is to have few experimental units without measurements to create homogeneous groups of measured observations. In a ranked set sampling, along with measured observations, H experimental units are selected at random from a population of interest. These are ranked from 1 to H based on some criterion such as researcher's opinion. Out the H units only one is measured and rest are assigned a position relative to the one measured. This process is repeated n times to get a ranked set sample. The judgment ranks act like a stratified sample and the RSS design can improve the efficiency. The major difference between the two approaches is that, in the RSS design ranking is done prior to the measurement of one unit and it is done after the measurement is performed in the JPS design. The measured ones cannot be separated from other sampled items in the RSS design and hence a separate method has to be developed for the analysis of such designs. For JPS on the other hand, the analysis can be performed treating it as a simple random sample (SRS). Also, JPS sample may be unbalanced for small sample sizes and that additional variation in the sample size vector makes the JPS sample less efficient compared with a RSS sample. The purpose of this study is to use a JPS sample to draw distribution free statistical inference on the parameter of interest such as the quantile of order p. (20 refs.)
展开▼