Skip to main content
. 2020 Jan 27;69(5):1016–1032. doi: 10.1093/sysbio/syaa002

Table 4.

Results of reference runs (four runs with four Metropolis-coupled chains each) and test runs (16 single-chain runs) for each of the six data sets

  R.20M R.10M   2 eSPR : 1 eTBR   2 pSPR1 : 1 pTBR1   2 pSPR2 : 1 pTBR2
Data set ASDSF   ASDSF Inline graphic   ASDSF Inline graphic   ASDSF Inline graphic
SQ10 0.004 0.007   0.097 4.56%, 3.78%   0.037 8.52%, 12.1%   0.012 8.63%, 12.2%
DA10 0.003 0.006   0.019 8.24%, 5.02%   0.016 8.11%, 7.04%   0.009 8.62%, 8.02%
LZ12 0.003 0.005   0.008 54.3%, 43.5%   0.006 48.2%, 43.1%   0.005 53.1%, 48.4%
CL12 0.004 0.006   0.009 18.0%, 12.7%   0.006 18.6%, 16.9%   0.006 19.8%, 18.8%
AZ12 0.004 0.007   0.031 4.88%, 4.18%   0.009 6.60%, 8.70%   0.017 7.14%, 10.0%
NP11 0.020 0.041   0.108 58.4%, 49.5%   0.039 46.9%, 44.1%   0.038 52.3%, 49.1%

Notes: We first give the average standard deviation of split frequencies (ASDSF) for the reference runs after 20 and 10 million generations (R.20M and R.10M, respectively). Then we give the ASDSF values and average acceptance proportion (Inline graphic) for 16 test runs using three different combinations of tree proposals. The reference tree samples after 20 million generations had ASDSF Inline graphic 0.02 and were considered as ground truth in the detailed studies of the convergence and mixing behavior of tree proposals.