Skip to main content

Table 1 Model comparison using code-length, or average negative cross-log-likelihood. Code lengths are given in bits. For Chromosome 7 data sets, SNP loci were sub-sampled in ENCODE regions to maintain an uniform coverage of the chromosome.

From: A model-based approach to selection of tag SNPs

 

Pop.(a)

SNPs(b)

Markov

HMM-2(c)

HMM-4D(d)

GR-1(e)

GR-2(f)

LS-HOM(g)

LS-HET(h)

Chr. 7

CEU

42835

–

13334.9

9737.9

10898.7

8273.4

6441.5

6030.6

Chr. 7

YRI

42790

–

16783.0

13996.9

15252.2

13543.4

9082.5

8705.6

ENr112

CEU

1134

378.6

238.4

152.8

127.2

89.2

53.6

51.4

ENr112

YRI

1082

489.9

348.9

238.9

225.0

157.8

80.7

77.9

ENr131

CEU

1188

454.0

306.7

161.5

151.7

100.2

66.1

60.2

ENr131

YRI

1080

439.7

331.9

244.2

227.4

172.4

101.9

92.7

ENr113

CEU

1375

478.7

287.8

160.4

120.6

88.8

57.9

55.9

ENr113

YRI

1525

597.9

424.9

286.3

228.3

157.5

84.9

81.7

ENm010

CEU

706

261.9

187.2

106.6

106.1

83.1

58.8

56.7

ENm010

YRI

741

325.5

250.5

175.4

177.8

152.1

106.6

101.4

ENm013

CEU

1001

417.5

279.6

132.2

83.7

57.1

38.2

37.6

ENm013

YRI

1111

452.1

336.3

211.0

157.0

108.7

63.2

62.0

ENm014

CEU

1110

442.1

290.1

140.9

104.6

71.5

55.3

50.9

ENm014

YRI

1224

483.2

338.9

237.0

166.5

117.7

71.7

68.4

ENr321

CEU

782

243.1

143.2

90.1

90.4

68.9

48.0

46.0

ENr321

YRI

1123

458.9

325.5

232.1

199.1

145.0

77.1

73.9

ENr232

CEU

627

189.4

117.2

89.9

98.5

82.3

64.1

58.4

ENr232

YRI

833

345.9

268.6

206.2

198.8

161.5

102.2

93.4

ENr123

CEU

1183

453.8

294.2

175.2

114.6

76.3

47.5

46.8

ENr123

YRI

1055

436.6

291.5

206.1

161.3

118.5

67.7

66.7

ENr213

CEU

800

323.9

207.1

87.4

82.3

59.5

41.0

38.5

ENr213

YRI

1085

418.5

319.9

219.8

178.7

131.6

75.4

71.6

  1. (a): Population. (b): Number of actually polymorphic SNPs in the population considered. (c-h): HMM-2 for the unconstrained two-state HMM; HMM-4D for the Daly et al. HMM; LS-HOM and LS-HET for the homogeneous and the heterogeneous Li and Stephens models; GR-1 and GR-2 for the "greedy" models with context sizes 1 and 2.