In article <199504022334.AA07604 at acacia.cbr.for.csiro.au>,
Charlie Bell <charlie.bell at cbr.for.csiro.au> wrote:
>>(in reply to Mark Siddall's:
>>Here's a piece of food for thought:
>>Consider having hit Joe's Zone due to a few highly deviant substitution
>>rates on some lineages/taxa.
>>How do we determine how stable these groupings are/are not.
>>I would argue that bootstrapping is irellevant here because
>>bootstrapping whacked out data will give you whacked out
>>pseudoreplicates.
...
>My first thought is to collect more data.
>Whether you use bootsrapping or any other method for error or reliability
>estimates, the results can only be as good as the data they are derived
>from. If your sample is not a good representation of the true population
>then you have a problem that statistics cannot overcome.
But as Mark notes this will not work. The point about an inconsistent
estimator is that it converges on the wrong answer with greater and
greater uncertainty as one collects more and more data. The only way out is
to use a method that isn't inconsistent (and there are such).
By the way, Mark's point is also made in my original 1985 paper.
Sorry for the slow response -- news reading was not fully working on a new
workstation.
-----
Joe Felsenstein, Dept. of Genetics, Univ. of Washington, Seattle, WA 98195
Internet: joe at genetics.washington.edu (IP No. 128.95.12.41)
