[phylocom-user] lttr interpretation

Adam Richman arichman at montana.edu
Fri Apr 2 10:00:25 PDT 2010


The phylocom function lttr compares the number of lineages through time 
for a given sample to that of a set of randomly generated samples of 
equal size. The rank position of the observed sample is determined by 
finding the total number of random samples which have fewer lineages for 
a given time interval. If none of the random samples have fewer, then 
the highest ranking is returned.

However, because of the way that lttr handles ties in determining rank 
position, the interpretation of the highest ranking is ambiguous.

To see this, consider the instance where, for a given time interval, the 
number of lineages in the observed sample equals the maximum possible. 
In this case, if any of the random samples also has the maximum number, 
then the highest ranking is returned.

Note that in this instance, the highest ranking has two very different 
possible explanations:

1. If a large proportion of the random samples also contain the maximum 
number of lineages, this indicates that the observed sample is not 
statistically different from random samples of the larger community.

2. If only a small fraction of random samples also contain the maximum 
number of lineages, then the observed sample is apparently quite 
different from the random samples, since it tends to have relatively 
divergent (older) lineages.

I suggest that when the maximum ranking is obtained, that the rank of 
the first tie which led to this maximum ranking also be reported. This 
further information can be used to distinguish the two cases.

Adam Richman



More information about the phylocom-user mailing list