Table 3.
Comparison of switch error rates.
| n | L | IBD | BGL | BOTH |
|---|---|---|---|---|
| 50 | 10 | 3.88% | 6.17% | 2.41% |
| 100 | 10 | 2.99% | 4.54% | 1.81% |
| 200 | 10 | 1.96% | 2.23% | 1.18% |
| 50 | 5 | 3.73% | 4.87% | 1.83% |
| 100 | 5 | 2.13% | 2.02% | 1.01% |
| 200 | 5 | 1.47% | 1.01% | 0.63% |
| 50 | 2 | 3.81% | 3.16% | 1.57% |
| 100 | 2 | 2.70% | 1.20% | 0.62% |
| 200 | 2 | 1.55% | 0.55% | 0.35% |
The table shows results for the switch error rate by, respectively, our phase update algorithm (see Appendix C for pseudocode) in the IBD column, the BEAGLE algorithm in the BGL column, and the combined approach in the BOTH column for a dataset simulated by using N = 10000 as the size of the original population, T = 100 as the number of generations for which the population was simulated, n as the size of the family sampled, t = 5 as the number of generations for which the family was simulated, L as the genetic length in centimorgans of the haplotypes simulated, and s = 400 as the number of SNPs in the haplotype.