Skip to main content

Table 3 Error rates found by each method when using cut-offs for the probabilities provided by PHASE

From: Computation of haplotypes on SNPs subsets: advantage of the "global method"

GH1 – cutoff 70% (5 SNPs)
MD Method Abs IER Rel IER Res rate
0% Global 2.18% 2.24% 97.31%
  Local 2.56% 2.65% 96.54%
2% Global 2.92% 2.99% 96.78%
  Local 2.97% 3.05% 94.78%
5% Global 3.33% 3.49% 95.38%
  Local 3.17% 3.51% 90.42%
10% Global 4.50% 4.92% 91.40%
  Local 4.07% 5.00% 81.50%
15% Global 5.83% 6.56% 88.88%
  Local 4.85% 6.42% 75.54%
20% Global 7.06% 8.14% 86.80%
  Local 6.00% 8.81% 68.12%
GH1 – cutoff 50% (5 SNPs)
MD Method Abs IER Rel IER Res rate
0% Global 2.99% 2.99% 99.82%
  Local 3.50% 3.50% 99.83%
2% Global 3.82% 3.98% 99.81%
  Local 4.18% 4.26% 97.47%
5% Global 5.03% 5.06% 99.42%
  Local 5.08% 5.32% 95.45%
10% Global 7.17% 7.28% 98.45%
  Local 7.24% 8.04% 90.06%
15% Global 9.41% 9.62% 97.84%
  Local 9.50% 10.95% 86.83%
20% Global 11.25% 11.58% 97.18%
  Local 11.96% 14.45% 82.77%
  1. This table presents the summary of average error rates when using a cut-off on resolution probability given by PHASE instead of Rmax when choosing the resolution obtained by each method. The 2 cut-offs chosen were respectively 50% and 70%. The error rate is almost always lower than with Rmax, but the number of assigned haplotypes strongly decreased.
  2. IER: individual error rate, Abs IER: absolute IER, Rel IER: relative IER, Res Rate: resolution rate, MD: Missing Data.