This article explores the efficiency and finite distribution of some easy confidence interval procedures for location parameter. A new confidence procedure based on few order statistics is proposed. The efficiency of the robustness of this confidence procedure is measured by comparing its expected length to its competitors. The simulated sample sizes are from 4 to 10 and distributions include the Gaussian and some moderately long-tailed distributions. The proposed procedure performs quite well in terms of the lengths of its confidence interval for small samples and high confidence levels. Finally, appropriate Student's t distributions are fitted to facilitate the use of the proposed procedure.
