The results of this analysis are shown in Figure 3. The x-axes in this figure display the discordance between imputed genotype calls and observed Illumina calls, which is a surrogate for imputation error rate; the y-axes display the percentage of genotypes for which no call was made. Each method's line is formed by considering several different calling thresholds for imputation posterior probabilities. For example, a certain number of maximum posterior probabilities will exceed a threshold of 0.9, and among these we can ask what percentage of the best-guess imputed genotypes disagree with the Illumina genotypes. This yields an x-coordinate, and the y-coordinate is simply the percentage of all imputed genotypes for which no posterior probability exceeds the threshold. We generated the lines on the plots by repeating these calculations for calling thresholds ranging from 0.33 to 0.99 for each method.
