Pemodelan Data Angka Kematian Bayi Menggunakan Regresi Robust

Abstract

The ordinary least square method (MKT) is a classic method to estimate parameters in regression modelling. The MKT has classical assumption, where these assumptions must be fulfilled. One of the causes of violation of classical assumptions is the presence of outlier observations. Robust regression is an alternative for correcting parameters in data that indicate outliers. In this article, we apply robust regression using estimation-M in data modelling of infant mortality rates that occur in West Java. Based on the mean squared error (MSE) criterion in the estimation-M, it is obtained that the Huber approach has a lower MSE than Tukey Bisquare. The results of estimation using Huber show that only the low number of births has a significant effect on infant mortality in West Java, with an effect of 2.047.