Skip to main content

Table 1 Rankings of methods for building a summary tree

From: Looking for trees in the forest: summary tree from posterior samples

Method

RH

CME

CCE

CAE

DVE

MF

TLL

CLL

TP(med)

1/3

0/0

12/9

8/8

7/5

3/3

0/0

3/3

TP(avg)

0/4

0/0

13/9

6/7

0/3

11/10

1/6

14/15

MED,TCB

1/0

3/3

10/7

6/4

6/6

9/7

8/11

9/9

MED,MCC

1/0

6/6

12/10

7/4

7/7

7/6

7/10

7/7

RBS,TCB

6/8

10/10

4/3

12/12

11/10

2/2

4/3

1/1

RBS,MCC

7/9

12/12

5/4

12/12

11/11

1/1

3/2

0/0

HSO,TCB

1/1

2/2

11/8

6/4

6/6

10/7

8/11

10/10

HSO,MCC

1/2

5/5

13/11

6/4

7/7

8/6

7/10

8/8

SRBS,TCB

3/5

8/7

8/5

6/5

1/2

12/9

6/9

13/13

SRBS,MCC

4/6

9/9

9/6

6/5

3/4

11/9

5/9

11/12

RAS,MCC

5/6

14/14

7/4

9/9

9/8

4/4

3/4

6/6

RAS,TCB

5/6

13/13

6/3

10/10

10/9

5/4

5/5

5/5

mSRBS

3/5

9/8

9/6

7/6

2/3

11/8

4/8

12/11

mRAS

5/7

15/15

7/4

11/11

10/9

6/5

6/7

4/4

mRBS

6/8

11/11

3/3

13/13

11/10

0/0

2/1

2/2

mHS

1/0

18/19

1/1

2/0

1/0

16/12

12/14

18/18

AVG,MCC

0/4

7/7

11/9

5/5

5/8

13/11

9/12

15/14

CAT,TCB

0/5

1/1

14/10

0/0

1/2

18/16

14/15

20/21

CAT,MCC

0/4

4/4

15/12

1/1

1/2

17/15

13/14

19/20

HS,TCB

2/2

17/17

2/2

4/3

4/1

15/14

11/17

16/16

HS,MCC

1/1

19/18

2/2

3/2

3/0

14/13

10/16

17/17

CONS(med)

1/0

16/16

0/0

3/2

8/8

17/14

15/13

19/19

  1. Rankings of methods for building a summary tree from posterior samples. Both the comparison and error magnitude ranking are given for each method and 7 error measures (as a comparison/magnitude pair). The error measures are root height error (RH), clades missed (CME), clades called (CCE), clade ages errors (CAE), divergence times errors (DVE), model fit (MF) and tree likelihood/coalescent likelihood (TLL/CLL). Method names are as defined in the methods section, except for CONS,MED,AVG and HSO. CONS is the strict consensus tree with ages set by median estimates, as implemented by DendroPy. MED and AVG respectively use the median and average of clades ages from all matching trees in the posterior. HSO also uses the same clade ages, but uses the search algorithm utilized by the tree distance methods to find heights which minimize the total squared error.