Fig. 3
From: Sparse Phenotyping and Haplotype-Based Models for Genomic Prediction in Rice
Genomic prediction accuracies of days to heading (DTH) and plant height (PH) in the a first, b second, and c third populations using three prediction models (VG, VGR, FA) with different missing rates of environment-specific best linear unbiased estimates (BLUEs) of genetic effects of lines in the training set. The genetic effect was respectively described by SNP genotypes (Marker-based), haplotype using complete blocks (Haplotype-based-block-complete), and short haplotypes containing two SNPs (Haplotype-based-block-2SNP) and three SNPs (Haplotype-based-block-3SNP) within a haplotype block. The whiskers at the top of the bars indicate the standard deviations of prediction accuracies in different cross-validation repeats. The asterisks above the bars indicate the prediction accuracies of haplotype-based approaches were statistically significantly (p < 0.05, t-test) higher (black) and lower (red) than those of the marker-based approach after a Fisher's z transformation