Mathematics · Statistics and Probability
A novel class of imputed estimators of population mean using robust quantile regression
Lovleen Kumar Grover, Anchal Sharma
See on the mapAbstract
Missing observations is a common issue in surveys for collecting data. The statistical inference is greatly affected due to non-response which is efficiently handled by imputation technique. In addition to non-response, data is often contaminated with outliers. Traditional Ordinary Least Square method is not an effective tool in curbing the negative impact of outliers and thus, robust methods must be used. Numerous ratio/regression type estimators of population mean of study variable have been suggested using robust techniques when complete information is available. A very less approach has been made when the data is affected both by non-response and outliers of study variable. This paper presents three novel robust quantile regression cum ratio type logarithmic estimators of population mean under the assumption that the information on conventional/non-conventional measures of location parameters of auxiliary variable is available. We derive the approximate expressions for the mean square errors of these classes. The optimum mean square errors of proposed estimators are compared with those of the existing estimators. A numerical investigation is used to validate the conclusions so acquired. The empirical results illustrate that the proposed estimators outperform their present counterparts across all the three strategies under consideration, as evidenced by consistently higher Percent Relative Efficiencies. The integration of quantile regression and logarithmic transformation results in notable efficiency benefits for all of the proposed estimators. This improvement demonstrates the effectiveness and dependability of quantile regression techniques, particularly when handling outliers or skewed data.