Validation and evaluation of P values from comparisons using 16S rRNA gene libraries from the studies of Bond et al. (3), Dunbar et al. (10), and McCaig et al. (20), using various testing conditionsa

Site (reference)Library name (no. of sequences)Published LIBSHUFF P valuebOur P value, using LIBSHUFFbCorrected LIBSHUFF P valuebP value, using ∫-LIBSHUFF for ΔD value ofb:P value for integral form
Sequencing batchSBR1 (97)0.3080.3000.0870.0760.0660.0650.065
    reactors (3)SBR2 (92)0.8240.9510.7980.8080.7440.7470.747
Arid soils (10)C0 (59)0.0420.0430.0110.0110.0120.0120.012
S0 (53)0.3980.2900.1710.1770.1880.1870.187
Scottish soil (20)SAF (138)0.1200.1390.0350.0320.0320.0300.030
SL (137)0.1350.2560.0740.0760.0740.0730.073
  • a For each pair of libraries, the first row of P values indicates when results when the first library is the homologous library, and the second row indicates results when the second library is the homologous library. The margin of error for the P value's 95% confidence interval when P values are near 0.05 for the comparison between the published P values of Singleton et al. (27) and our implementation of the uncorrected and corrected LIBSHUFF was approximately 0.014 (1,000 randomizations). For the simulations performed using ∫-LIBSHUFF, it was approximately 0.004 (10,000 randomizations).

  • b The range of summation was between 0.00 and 0.50.