Tree comparison p-KH=0 for the best tree?

[Charles Pouchon]

Hi all, 

I'd be grateful if you could clarify a point about the topology tests in IQ-tree: 

I am using this test (in IQ-tree v.2.1.2) to compare 5 trees: (1) the ML unconstrained tree generated from IQ-tree, a (2) bayesian tree, a (3) MP tree and two trees (4, 5) with custom constrained topologies. 

I get the following result: 

Tree      logL    deltaL  bp-RELL    p-KH     p-SH       c-ELW       p-AU

-------------------------------------------------------------------------

  1 -124988.3788       0   0.254 +      0 -  0.799 +      0.27 +    0.678 + 

  2 -124988.3788       0   0.261 +      0 -  0.799 +      0.27 +    0.678 + 

  3 -124989.7063  1.3275   0.439 +  0.441 +  0.602 +     0.416 +    0.477 + 

  4 -125003.8226  15.444  0.0445 - 0.0822 +  0.171 +    0.0412 -   0.0792 + 

  5 -125007.2398  18.861  0.00147 - 0.0401 -  0.106 +   0.00238 -  0.00206 - 

The (1) ML and the (2) bayesian tree are identical (=same topology). However, I don't understand why I have a p-KH of 0 for these two trees (1) and (2). 

Do know why IQ-tree returns a zero p-value for both trees? 

Thank you very much. 

Best regards, 

Charles

[Minh Bui]

Hi Charles,

Interesting, I didn’t expect this. Except for p-KH, the other test results look reasonable.

Is this problem gone if you remove tree 2 from the test? (as it’s topologically identical to tree 1 anyway).

On a side note, please never use KH test for multiple testing: it’s only designed for testing 2 trees. This test is in IQ-TREE for historical reason only.

Cheers,

Minh

--
You received this message because you are subscribed to the Google Groups "IQ-TREE" group.
To unsubscribe from this group and stop receiving emails from it, send an email to iqtree+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/iqtree/f3771501-40c1-4af5-a85e-b9caaafc3241n%40googlegroups.com.

[Charles Pouchon]

Hi Minh,

Thank you for you response. Maybe such issue arises when using same topologies. But If I used the weighted KH test, it doesn't return a zero value. 

I have removed the second tree to keep a single topology and I no longer have this issue. 

Tree      logL    deltaL  bp-RELL    p-KH     p-SH    p-WKH    p-WSH       c-ELW       p-AU

-------------------------------------------------------------------------------------------

  1 -124988.3512       0   0.518 +  0.564 +      1 +  0.564 +  0.844 +     0.515 +     0.68 + 

  2 -124989.6892  1.3381   0.441 +  0.436 +  0.603 +  0.436 +  0.593 +     0.441 +    0.478 + 

  3 -125003.8001  15.449  0.0388 - 0.0782 +  0.166 + 0.0782 +  0.147 +    0.0408 -   0.0735 + 

  4  -125007.188  18.837  0.0019 - 0.0386 -  0.103 + 0.0386 - 0.0976 +   0.00314 -  0.00137 - 

If the two toplogies are identical, it's true that I don't need to consider both trees in this analysis. 

And thanks for your advice for the KH test, I'll take it off from my analysis. 

Cheers,

Charles

[Minh Bui]

Hi Charles,

Yes, if two trees have identical topology you just need to put one into the test.

Mi

To view this discussion on the web visit https://groups.google.com/d/msgid/iqtree/0274121e-ba5c-4f5b-a578-6799ea8c002fn%40googlegroups.com.

[Charles Pouchon]

Hi Minh, 

It is what I have done now :) 

Thank you. 

Charles