Indonesian Journal of Applied Statistics Volume I No.1 May 2018 ISSN. 2621-086X 37 A Robust Regression by Using Huber Estimator and Tukey Bisquare Estimator for Predicting Availability of Corn in Karanganyar Regency, Indonesia Hasih Pratiwi 1 , Yuliana Susanti 2 , and Sri Sulistijowati Handajani 3 1,2,3 Statistics Study Program, Universitas Sebelas Maret, Jl. Ir. Sutami 36A, Surakarta 57126, Indonesia 1 hpratiwi@mipa.uns.ac.id 2 yulianasusanti@staff.uns.ac.id 3 rr_ssh@staff.uns.ac.id Abstract. Linear least-squares estimates can behave badly when the error distribution is not normal, particularly when the errors are heavy-tailed. One remedy is to remove influential observations from the least-squares fit. Another approach, robust regression, is to use a fitting criterion that is not as vulnerable as least squares to unusual data. The most common general method of robust regression is M-estimation. This class of estimators can be regarded as a generalization of maximum-likelihood estimation. In this paper we discuss robust regression model for corn production by using two popular estimators; i.e. Huber estimator and Tukey bisquare estimator. Keywords : robust regression, M-estimation, Huber estimator, Tukey bisquare estimator 1. Introduction Corn is one of the important food crops besides rice and wheat. Some people in Indonesia such as in Madura use corn as a staple food which has advantages and benefits as the highest source of carbohydrates [1,4,5,12]. Because of the importance of addressing the needs of the food, we require an effort to predict production in the future. There are several methods that can be used to predict corn production as well as to investigation several factors that influence it, such as regression analysis [6,10]. The problems that arise in the regression analysis is to determine the best estimators for model parameters, which is heavily influenced by the use of the method. For example, using the least squares method would not be appropriate in solving problems contains several outliers or extreme observations, or the assumption of normality can not be met. By using regression analysis, the production of which go far beyond the production can generally be categorized as an outlier, so using the least squares method to estimate the regression parameters is less precise [2,13]. To overcome this problem, we require a parameter estimation method which is robust. Robust interpreted as insensitivity or resilience to small changes of assumptions. Estimation using the maximum likelihood estimate (MLE) will produce an estimator of the same