TABLE 4

Estimated P values for the three F-statistic-based tests of the null hypothesis of constant population size for various levels of variability in μ

Level of variability in ^
NoneModerateMediumHigh
Data setPopulationgF1F2gF1F2gF1F2gF1F2
FirstLuo0.810.160.140.510.060.050.320.030.020.120.000.00
Sardinia0.770.830.810.460.540.500.270.340.310.080.140.11
Kaingang0.810.490.400.510.230.180.320.100.070.120.010.01
Pooled0.700.250.210.390.100.080.210.040.030.050.000.00
SecondAfrica0.750.740.720.450.450.420.270.260.240.090.080.07
Asia0.910.340.290.660.150.120.500.070.050.310.010.00
Europe0.930.110.080.700.040.030.530.010.010.350.000.00
Sardinia0.910.060.040.670.020.010.490.010.000.310.000.00
Pooled0.840.250.220.570.110.090.380.040.030.170.000.00
First*Luo0.820.200.170.530.080.060.350.030.020.140.000.00
Sardinia0.830.070.050.530.020.020.350.010.000.140.000.00
Kaingang0.840.350.260.550.160.110.370.060.040.160.010.00
Pooled0.750.020.020.450.010.000.250.000.000.080.000.00
  • Significance levels of tests of the null hypothesis of constant population size based on three test statistics, g (effectively introduced by Reich and Goldstein 1998), which uses only population variance at each locus, and F1 and F2, introduced in this article, based on NPV at each locus. The first and first* data sets refer to the data reported in Di Rienzo et al. (1998), including and excluding, respectively, the locus D19S244. The second data set refers to that reported in this article. Significance levels are given under a range of assumptions about the extent to which the mutation rate varies across loci: no variability, moderate variability (5% of loci having mutation rates 10−3 and 10−5, respectively, with the remainder having rate 10−4), medium variability (10% of loci having mutation rates 10−3 and 10−5, respectively, with the remainder having rate 10−4), and high variability (20% of loci having mutation rates 10−3and 10−5, respectively, with the remainder having rate 10−4). Significance levels are based on 30,000 values simulated under the null hypothesis, with N = 10,000. (Changes in the value of Nin the range 5,000 to 20,000 have little effect on the significance level; data not shown.) As described in appendix b, tests based on F1 and F2 are shown to be conservative under variation in the mutation mechanism across loci and error in the estimation of mutation mean square.