I wonder if someone could advice or point a direction in trying answering
hypothesis testing-like questions in a tree problem.
Suppose I have two hierarchical trees organising people according to their
music preferences and to some social interaction measure.
Each person/ leaf has one extra character which can take values A, B and C
(e.g. parent's supportiveness on musical education).
I wanted to select people from A and test across several tree distances if
the friends they keep between two trees tend to be more A or C then the
friends from say group B. So testing homophily of characters through the
tree structure. The goal of the toy example is to discover if the character
(parent's supportiveness) could be driving the transitions/changes between
the two trees.
I was wondering if a Pearson's chi-square test/ contingency table would be
appropriate in this scenario. As I see it, there are two cohorts:
Cohort1: people from group A. Cohort2: people from group B.
Then, for a given tree distance (dist=2,3,4 ... simply number edges between
leaves), test for instance whether a character (say A) of common friends
between trees two trees depends or not on picked leaf being A.
Questions:
1) is this sensible?
2) Possible problems: double counting (when I count if friend x is A of a y
person A, then when the friend y of x will also be counted) and
dependencies through the tree (lost friends at tree distance 4 may be
actually a new friend at distance 2)
3) the trees were constructed manually into an newick format without edge
lengths. I am currently assuming all lengths equal to 1 (topology only). Is
this ok? Does anybody know of an automated way of getting edge lengths
which are proportional to node's depth. Ape R package dist.nodes and
cophenetic give really odd values.
Many thanks!
Inigo
[[alternative HTML version deleted]]
_______________________________________________
R-sig-phylo mailing list - R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
