prior range of substitution rates

[sergios-orestis kolokotronis]

In the absence of reliable calibrations, one can use a fixed
substitution rate (not the best, since it's an imported rate from
other studies) to estimate taxon divergence times. Is there a way to
set a a [e.g. uniform] prior on the rate within logical bounds, i.e.
not orders of magnitude? A simple way is to run the analysis with
several fixed rates, but I was just wondering.

Thanks
sergios

[Andrew Rambaut]

Dear Sergios,

If you don't fix the substitution rate, you can put a prior
probability distribution on it in the Priors panel. I suggest you
don't use a uniform but try to provide a distribution like the normal
with a mean and variance that reflect the uncertainty in the rate
estimated by the other studies.

Andrew

___________________________________________________________________
Andrew Rambaut
Institute of Evolutionary Biology University of Edinburgh
Ashworth Laboratories Edinburgh EH9 3JT
EMAIL - a.ra...@ed.ac.uk TEL - +44 131 6508624





--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.

[alexei]

I would recommend lognormal rather than normal. The normal
distribution supports values below 0, which makes no sense for a scale
parameter like substitution rate. A log normal with appropriate mean
and 95% HPD bounds is a good choice for substitution rate prior in my
opinion. If you have a number of *independent* rate estimates from
other papers, you can simply fit a log normal distribution to them as
use that as a prior for your new analysis.

Cheers
Alexei

On Sep 25, 11:40 pm, Andrew Rambaut <a.ramb...@ed.ac.uk> wrote:
> Dear Sergios,
>
> If you don't fix the substitution rate, you can put a prior  
> probability distribution on it in the Priors panel. I suggest you  
> don't use a uniform but try to provide a distribution like the normal  
> with a mean and variance that reflect the uncertainty in the rate  
> estimated by the other studies.
>
> Andrew
>
> On 25 Sep 2009, at 11:00, sergios-orestis kolokotronis wrote:
>
>
>
> > In the absence of reliable calibrations, one can use a fixed
> > substitution rate (not the best, since it's an imported rate from
> > other studies) to estimate taxon divergence times. Is there a way to
> > set a a [e.g. uniform] prior on the rate within logical bounds, i.e.
> > not orders of magnitude? A simple way is to run the analysis with
> > several fixed rates, but I was just wondering.
>
> > Thanks
> > sergios
>
> ___________________________________________________________________
>   Andrew Rambaut
>   Institute of Evolutionary Biology       University of Edinburgh
>   Ashworth Laboratories                         Edinburgh EH9 3JT

>   EMAIL - a.ramb...@ed.ac.uk                TEL - +44 131 6508624      

[Denzel]

Thanks for the suggestions. But what about situation where rate
estimated from other studies are not avaiable. It seems a hyper-prior
for the log-normal is desired, where the parameters for log-normal are
estimated/(or integrated out), i.e, [Suchard 2003]. I am not sure if
it is supported in BEAST.

[Suchard 2003]: Suchard MA, Kitchen CMR, Sinsheimer JS and Weiss RE.
Hierarchical phylogenetic models for analyzing multipartite sequence
data, Systematic Biology 52:649-664, 2003

> > Scotland, with registration number SC005336.- Hide quoted text -
>
> - Show quoted text -