A classification algorithm, based on a multi-chip, multi-SNP approach for Affymetrix SNP arrays. Using a large training sample where the genotype labels are known, this aglorithm will obtain more accurate classification results on new data. RLMM is based on a robust, linear model and uses the Mahalanobis distance for classification. The chip-to-chip non-biological variation is removed through normalization. This model-based algorithm captures the similarities across genotype groups and probes, as well as thousands other SNPs for accurate classification. NOTE: 100K-Xba only at for now.