Skip to content

Conversation

@stelzch
Copy link

@stelzch stelzch commented Jan 16, 2025

It seems the overall MSA length is used as sample size when computing the information criteria. In partitioned analyses, the MSA length does not coincide with the partition length.

Difference in AICc and BIC score can be observed in the following example:

partitions.txt:

DNA, p1 = 1-400
DNA, p2 = 401-2000
DNA, p3 = 2001-14000
DNA, p4 = 14001-14043

modeltest-ng -i example-data/dna/birds.fas -q partitions.txt

Without change applied
Partition 1/4:
                         Model         Score        Weight
----------------------------------------------------------
       BIC              HKY+G4     5153.5607        0.7914
       AIC           TPM3uf+G4     4999.0789        0.1581
      AICc           TPM3uf+G4     4999.0789        0.1581

Partition 2/4:
                         Model         Score        Weight
----------------------------------------------------------
       BIC           TIM2+I+G4    17336.8660        0.9997
       AIC           TIM2+I+G4    17163.2188        0.6466
      AICc           TIM2+I+G4    17163.2188        0.6466

Partition 3/4:
                         Model         Score        Weight
----------------------------------------------------------
       BIC            GTR+I+G4   143836.0189        0.9590
       AIC            GTR+I+G4   143647.2719        0.9990
      AICc            GTR+I+G4   143647.2719        0.9990

Partition 4/4:
                         Model         Score        Weight
----------------------------------------------------------
       BIC              K80+G4      633.6630        0.6091
       AIC             TPM2+G4      502.8833        0.0908
      AICc             TPM2+G4      502.8833        0.0908
With change applied
Partition 1/4:
                         Model         Score        Weight
----------------------------------------------------------
       BIC              HKY+G4     5082.3924        0.3576
       AIC           TPM3uf+G4     4999.0789        0.1581
      AICc           TPM3uf+G4     5001.0789        0.1631

Partition 2/4:
                         Model         Score        Weight
----------------------------------------------------------
       BIC           TIM2+I+G4    17286.9072        0.9975
       AIC           TIM2+I+G4    17163.2188        0.6466
      AICc           TIM2+I+G4    17163.2188        0.6466

Partition 3/4:
                         Model         Score        Weight
----------------------------------------------------------
       BIC            GTR+I+G4   143832.0885        0.9620
       AIC            GTR+I+G4   143647.2719        0.9990
      AICc            GTR+I+G4   143647.2719        0.9990

Partition 4/4:
                         Model         Score        Weight
----------------------------------------------------------
       BIC             TPM2+G4      534.5849        0.2349
       AIC             TPM2+G4      502.8833        0.0908
      AICc              K80+G4      529.3150        0.4007

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant