On some data oriented robust estimation procedures for means

Data oriented to estimate means is very important for large data sets. Since outliers usually occur, the trimmed mean is a robust estimator of locations. After building a reasonable linear model to explain the relationship between the suitably transformed symmetric data and the approximately standardized normal statistics, we find the trimmed proportion based on the smallest variance of trimmed means. The related statistical inference is also discussed. An empirical study based on an annual survey about inbound visitors in the Taiwan area is used to achieve our goal in deciding the trimmed proportion. In this study, we propose a complete procedure to attain the goal.