Yule and Birth-Death models: units="substitutions" or "generations" or "years"

[Jessie]

Hello Beast users,

would anyone be able to clarify whether it matters how the units are
defined in the yule or birth-death model elements?

i.e. as units="substitutions", units="generations" or units="years"

I am estimating a phylogeny for 200 plant species, with 37 calibration
densities on tMRCAs, in units of Myr.

I am uncertain about whether I should modify the XML for the Yule or
birth-death model elements to give units="years", (as I have for the
starting tree), or if this is a redundant thing to do, if the analysis
also includes calibrations.

BACKGROUND:

The <yuleModel> element example in the XML definition begins:
<yuleModel conditionalOnRoot="true" units="generations">
(URL: http://beast.bio.ed.ac.uk/BEAST_v1.7.x_XML_Reference#.3CyuleModel.3E_element
)

and similarly for the <birthDeathModel> element:
<birthDeathModel type="foo" conditionalOnRoot="true"
units="generations">

However, the output when selecting either model in BEAUti
automatically gives: <.... units="substitutions">
Similarly, the example XMLs for calibration studies (e.g.
primates.xml) also state: <... units="substitutions">

The BEAST website FAQ explains that the prior on evolutionary rates
(or any calibrations, or dated tips) determines whether the resulting
trees are in units of time or substitutions, and also affects the
demographic parameters (which would logically also include the birth /
death rates for species tree models? )
http://beast.bio.ed.ac.uk/FAQ#What_units_are_mutation_rates.2C_root_heights_and_population_sizes_expressed_in.3F

Similarly, under the description of parameters:
" yule.birthRate This parameter is the rate of lineage birth in the
Yule model of speciation. If clock.rate is 1.0 then this parameter
estimates the number of lineages born from a parent lineage per
substitution per site. If the tree is instead measure in, for example,
years, then this parameter would be the number of new lineages born
from a single parent lineage per year."

All of which suggests to me that it doesn't matter that the birth /
death rates are initially stated with units="substitutions" or
"generations" or "years", but it seems like an important thing to be
sure about.

Thanks very much for any responses
- Jessie

[Jessie]

Following several responses from other interested Beast users, my
conclusion from running models with alternative specification of units
in the Yule or Birth Death model elements, as:

units="substitutions", units="generations" or units="years"

-> the alternative declaration of units make no difference to the
results.
If you have supplied temporal information, then the units will all be
in Yrs (or Myrs or whatever time unit of the temporal information
itself), regardless of what the tree prior element says.


It would be great if this point could be clarified in the XML
specification, or the FAQ.

Otherwise, the inclusion of redundant or incorrect information going
into the XML, is inconsistent with the philosophy of the XML providing
an exact and comprehensible specification of the model.
From the perspective of BEAST users it is confusing and makes it very
hard to know which details deserve attention and which should be
ignored..

thanks
- Jessie.

On May 8, 8:42 pm, Jessie <jessie.wel...@gmail.com> wrote:
> Hello Beast users,
>
> would anyone be able to clarify whether it matters how theunitsare
> defined in the yule or birth-death model elements?
>

> i.e. asunits="substitutions",units="generations" orunits="years"

>
> I am estimating a phylogeny for 200 plant species, with 37 calibration

> densities on tMRCAs, inunitsof Myr.

>
> I am uncertain about whether I should modify the XML for the Yule or

> birth-death model elements to giveunits="years", (as I have for the

> starting tree), or if this is a redundant thing to do, if the analysis
> also includes calibrations.
>
> BACKGROUND:
>
> The <yuleModel> element example in the XML definition begins:
> <yuleModel conditionalOnRoot="true"units="generations">

> (URL:http://beast.bio.ed.ac.uk/BEAST_v1.7.x_XML_Reference#.3CyuleModel.3E_...

> )
>
> and similarly for the <birthDeathModel> element:
> <birthDeathModel type="foo" conditionalOnRoot="true"units="generations">
>
> However, the output when selecting either model in BEAUti
> automatically gives:    <....  units="substitutions">
> Similarly, the example XMLs for calibration studies (e.g.

> primates.xml) also state: <...units="substitutions">

>
> The BEAST website FAQ explains that the prior on evolutionary rates
> (or any calibrations, or dated tips) determines whether the resulting

> trees are inunitsof time orsubstitutions, and also affects the

> demographic parameters (which would logically also include the birth /

> death rates for species tree models? )http://beast.bio.ed.ac.uk/FAQ#What_units_are_mutation_rates.2C_root_h...

>
> Similarly, under the description of parameters:
> " yule.birthRate   This parameter is the rate of lineage birth in the
> Yule model of speciation. If clock.rate is 1.0 then this parameter
> estimates the number of lineages born from a parent lineage per
> substitution per site. If the tree is instead measure in, for example,
> years, then this parameter would be the number of new lineages born
> from a single parent lineage per year."
>
> All of which suggests to me that it doesn't matter that the birth /

> death rates are initially stated withunits="substitutions" or