Executive Summary : | The inferential procedures are generally presume that the survey data are free from the measurement errors (ME) which may not be justifiable in real life and the data may incorporate ME because of several inevitable causes. The ME is defined as the difference between the true and observed values of the variable. In any survey data, the ME are an essential part of errors. These errors occur in the process due to various sources which are practically unavoidable but the proper caution during the estimation procedures helps to reduce the impact of ME so that their existence may still be statistically valid. However, circumstances may customarily arise when the presence of ME frequently alter the conclusion. The studies conducted till date are done under the supposition that the ME are uncorrelated. This is more of a mathematically simplifying assumption which resolves itself to neat and clean solutions and results. While, in practice, the data on the variables under consideration are beset with correlated ME. Therefore, it is more logical to consider the case where ME in the variables are correlated. McIntyre's ranked set sampling (Rss) is found to be more efficient that simple random sampling. A Rss scheme is applicable in the following situations: 1. The ranking of a set of sampling units can be done easily by judgment. 2. There are certain easily obtainable concomitant variables available. The proper use of Rss can lead to substantial gains in efficiency but the problem of non-response along with the correlated measurement errors can impair the related inferences of the parameters. Therefore, it is very crucial to propose robust inferential procedures which may still provide the best possible results under such unpleasant but realistic scenarios. This study precisely wishes to achieve this objective which is elaborated point wise in the next section. |