Skip to main content

Table 4 Success rate (%) of data analysis methods for a number of species ranging from two to five; mtDNA sequences were simulated for a reference sample size n = 10 and a separation time T = 500 (N f /2); the mutation parameter θ was either 3 (A) or 30 (B).

From: DNA barcode analysis: a comparison of phylogenetic and statistical classification methods

Number of species NJ PhyML 1-NN CART RF Kernel P < 0.05 P < 0.01
(A) θ = 3         
2 † 87.25 86.30 87.30* 87.15 87.20 87.15   
3 81.73 * 80.77 80.67 80.40 80.97 81.10   
4 75.80 75.00 75.40 75.68 75.95 * 74.78 ¶ 1  
5 73.26 * 72.36 72.58 72.84 73.22 70.74 ¶ 1 1
(B) θ = 30         
2 † 96.10 96.20 * 95.55 93.50 ¶ 95.25 94.00 ¶ 2 2
3 94.40 * 94.23 94.00 90.93 ¶ 93.50 92.10 ¶ 2 2
4 93.78 * 93.73 92.90 90.10 ¶ 92.53 ¶ 91.40 ¶ 3 2
5 92.46 92.38 92.70 * 88.98 ¶ 92.08 90.46 ¶ 2 2
  1. * Best score
  2. ¶significantly below the best score. Columns 8 and 9 indicate the number of methods with p-values below 0.05 and 0.01 respectively.
  3. † Focal set of parameters, for comparison across tables.