Table 3 Positive predictive values from naïve Bayes analysis.

From: Differentiation of Hispanic biogeographic ancestry with 80 ancestry informative markers

Known Origin

SNP Panel

5 sets of 500 micro-simulations

347 real 1000 G genotypes

Sen. (%)

Spe. (%)

PPV (%)

Sen. (%)

Spe. (%)

PPV (%)

CLM

Setser80

81.9%

70.1%

47.8%

17.0%

98.4%

80.0%

Seldin96

84.2%

77.9%

55.9%

55.3%

90.9%

69.3%

Kidd44

63.2%

49.9%

29.6%

51.1%

83.8%

53.9%

PUR

Setser80

90.4%

86.9%

69.8%

81.7%

85.2%

70.2%

Seldin96

89.8%

80.5%

60.6%

89.4%

86.8%

74.4%

Kidd44

75.8%

51.7%

34.4%

71.2%

85.2%

67.3%

PEL

Setser80

98.1%

97.1%

91.8%

62.4%

97.7%

89.8%

Seldin96

99.4%

98.9%

96.9%

87.1%

95.4%

86.0%

Kidd44

91.8%

90.4%

76.1%

75.3%

94.7%

82.1%

MXL

Setser80

89.8%

79.5%

59.3%

84.4%

67.1%

36.7%

Seldin96

96.3%

92.0%

80.1%

51.6%

90.1%

54.1%

Kidd44

85.3%

68.3%

47.3%

50.0%

86.6%

45.7%

  1. Sensitivity, specificity, and positive predictive values from naïve Bayes leave-one-out cross-validation for the average of five sets of 500 micro-simulations (left) and n = 347 actual 1000 G genotypes (right). Micro-simulations were generated based on the allele frequencies from the 1000 G dataset only. Abbreviations used: Sen. = sensitivity, Spe. = specificity, PPV = positive predictive value, CLM = Colombia from Medellin, PUR = Puerto Rico, PEL = Peru from Lima, and MXL = Mexicans living in Los Angeles.