PERBANDINGAN ANALISIS DISKRIMINAN LINIER KLASIK DAN ANALISIS DISKRIMINAN LINIER ROBUST (Studi Kasus : Konsumsi Protein Berdasarkan Wilayah Perkotaan dan Pedesaan di Indonesia Tahun 2018)

Discriminant analysis is a statistical method used to classify an individual or object into a predetermined group based on the independent variables. In classical linear discriminant analysis, there are two assumptions that must be fulfilled, namely the independent variables with a multivariate normal distribution and the covariance matrix of the observed groups that are the same. Outlier data is observational data that is extreme from other observations. Therefore we need a robust estimator so that the discriminant analysis remains optimal even though there are data containing outliers. The MCD method is not widely used in dealing with outliers data. To evaluate errors in classification, the APER (Apparent Error Rate) value is used. APER value to represent the value of the sample proportion that is misclassified by the classification function. Discriminant analysis in this study aims to see the comparison of classical linear discriminant analysis and robust discriminant analysis. The research method is quantitative with 13 variables, divided into two categories, namely urban and rural areas. The analysis used is the classical assumption test. This will be applied to data on average protein consumption by food group by area of residence in Indonesia in 2018. The total result of the proportion of classical linear discriminant analysis is 10.45 percent and the total result of the proportion of robust linear discriminant analysis is 5.98 percent