Support

SUPPORT
and
RESAMPLING
Thanks to Leandro Gaetano
Support
Branch length
Bremmer support
Bootstrap
Jacknife
Support
Bremer Support (decay
index) (Bremer 1992) :
is the number of extra steps
needed to "collapse" a branch
the absolute amount of favourable evidence
supporting the group
Having a Bremer Support of 2 can have two meaning:
• A branch is supported by two uncontradicted characters. Therefore,
it can only “cost” two steps not to have that branch.
• A branch is supported by n characters, but contradicted by n-2.
Therefore, to prefer the contradictory branch “costs” only 2 steps.
Support
Support
Bremer support
Support
Bremer Support
BS= 2
BS= 1 There are some incongruences
between characters. Bremer
support will be the difference
between the number of
congruent characters (F) and
the number of characters that
contradict the branch
Support
Bremer Support: Problems
• If heuristic search is not complete and memory in insufficient (= hold
low), the consensus for the suboptimal trees can be overresolved
and the BS being artificially high
solution?
•What BS is considered good?
As a rule of thumb, a Bremer score of 3 is good and a score of 5
suggests that the group is highly supported
Support
relative Bremer Support (Goloboff and Farris, 2001)
Bremer relative = (F-C)/F
F = number of characters congruent with the branch
C = number of characters contradicting the branch
If relative Bremer support is 1 is totally supported if 0 not
supported at all
Resampling
Bootstrap
The bootstrap intention is to examine the relative
"confidence" associated with portions of a cladogram
in relation to character sampling (Sidall, 2001).
Resampling characters from the
original data matrix, creating a
pseudoreplicate matrix and then
recalculating the most parsimonious
tree on the pseudoreplicate
data.
From Soltis and Soltis (2003)
Stability can be defined only by reference to some factors (e.g., a group
stable under addition of characters may be very unstable under addition of
taxa or under recoding of some characters).
Support depends exclusively on presently available evidence (and, of
course, assumptions or theories used to interpret that evidence).
Resampling evaluates support because the frequency with which replicates
display a given group will be determined by the relative amounts of
favourable and contradictory evidence
From Soltis and Soltis (2003)
Resampling
Bootstrap: Problems
• Boostrap not valid statistically for very large data-matrices
• Unilateral support, meaning:
Recovered groups will be well supported
BUT might be not recovered groups that can have also good
support
• It is not clear what value of Boostrap is significant
Resampling
Jackknife
What would have happened if I
had fewer taxa or characters
than I presently have, and then
added the remainder?
• TAXONOMIC JACKKNIFE
• CHARACTER OR PARSIMONY JACKKNIFE
TAXONOMIC JACKKNIFE
1) Pseudoreplicas: remove one taxon from the data matrix and
search for most parsimonious tree(s). This is repeated
removing all the taxon from the matrix one every time
(Pseudoreplica include all taxa -1)
2) Consensus from the trees obtained with all the pseudoreplicas
3) The Jackknife value represents the proportion of trees in which
the groups recovered in the shortest tree(s) are represented
TAXONOMIC JACKKNIFE
From Sidall 2001
Bremer Support and Jackniffing can give different support values to monophyletic
groups.
Compare values between monophyletic groups LM and FG.