[Mesquitelist] S-value

[Jeffrey C. Oliver]

Hi Monica,

First question: With this approach, if your observed statistic falls within
the 95% CI, you cannot reject a hypothesis being tested.  Thus, for your
examples, you would not reject models 3 or 4.  However, the distribution of
observed values will depend on the number of replicates you are running.
Are you performing at least 100 replicates?  If possible, you could try
simulating 1,000, 10,000, or even more gene trees to get a better reflection
of the true distribution of your test statistic.

Second question: Best if your data reflect population-level sampling, thus
sequence (individual) trees are more appropriate, even if many individuals
have identical haplotypes.

-Jeff

On Mon, Feb 2, 2009 at 3:14 AM, Monica Guerrini <mguerrini at biologia.unipi.it
> wrote:

> Hi!
> Thank you very much for your answers to my
> question some days ago! Now I have resolved my
> first problem, that is importing and using a gene
> tree reconstructed with other programs! It is very easy!!!!
> But now I have another two questions.
> First of all if I measure the distributions of
> test statistics (Slatkin & Maddison's S) on a
> simulated population tree, when I reject or
> accept this model drawn? I know that if the
> observed S.-value of my gene tree falls outside
> of the 95% confidence interval, that model are
> not supported by the data. On the contrary, if I
> have a case similar to the example of Mesquite's
> manual I don't reject the model (S = 9, observed
> values within the 95% confidence interval, that is 4 and 10).
> But now I suppose this example:
> I measure the observed S-value for my first gene
> tree and that is 9 (for example...!).
> Now I draw some population trees corresponding to
> the evolution models I want to test. I measure
> the distribution of S-values for each model.
> model n° 1: the S-value distribution has only one
> value, for example 2, so I reject the model (outside the boundary)
> model n°2: the S-value distribution has two
> limits: 5 and 15, so I accept the model (inside the boundary)
> model n°3: the S-value distribution has only one
> value, that is 9 (exactly my observed S-value)      ??????
> model n°4: the S-value distribution has two
> limits: 9 and 15 (lower tail exactly my observed S-value)      ??????
> What happen for models n° 3 and 4? I'm not sure
> to accept or reject these hypothesis.......
>
> The second question is:
> For use a gene tree (reconstructed with other
> software) and to measure S-value it is better to
> use a sequences tree or an haplotypes tree?
>
> Thank you very very much again for your help!!!!!
>
>
> Monica
>
>
>
>
>
>
> _______________________________________________
> Mesquitelist mailing list
> Mesquitelist at mesquiteproject.org
> http://mesquiteproject.org/mailman/listinfo/mesquitelist
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mesquiteproject.org/pipermail/mesquitelist/attachments/20090202/25691d2b/attachment.html