Calibration file

[alexandre pedro]

Hello all,

I was trying to understand the calibration input file for DPPDIV but without success. Can anyone help me to get more of this information? After setting the number of calibration points and the offset exponential with min age for the root (I assume its in million of years, please correct me if I'm any wrong), how exactly do I specify the nodes I want to calibrate? Let's say I have an entire clade sister to a single branch, and this node has a fossil date:

tree: ((((a,b)(c,d)e),f),g)

and the fossil is between taxon g and the remaining a-to-f?

Thanks in advance for this immense help.

May I insert a pertinent question in the topic: why in DPPDIV, fossil calibrations are optional? I apologise if the question is too basic in any sense, but it isn't that trivial for me. Many thanks again.

Alex

[Tracy Heath]

Hi Alexandre,

The calibration nodes are referenced by any 2 tip names that descend from the left and right daughter lineages of the node. The tree you provided had a couple problems, primarily, this tree is not fully bifurcating (there is a polytomy) this is not acceptable for dppdiv, as it expects a binary tree. So I have changed the newick string:

(((((a,b),(c,d)),e),f),g);

Since the node you refer to the MRCA of g and a-f is the root of the tree, you can simply specify the node like this:

-U   root   100.2   157.3

This is a uniform calibration on the root. Alternatively you can specify 2 taxa for this same node:

-E   g   a   100.2  -m 140.0

This sets a uniform calibration density, offset by the fossil age "100.2" with a mean of 140. The ages can be proportional to millions of years.

Calibration is optional because you can still estimate relative divergence times without any calibrations. Good calibrations are rare for many groups, but estimates of relative node ages are all that is needed for answering some questions like understanding the relative rates of diversification  or the evolution of continuous traits. Without fossils, the node ages are all reported as ages relative to the root. 

Let me know if you have any further questions!

Cheers!

Tracy

[Luis]

Hi Tracy,

I am using FDPPDIV for the first time. I'm trying to set up a file containing the calibration nodes of my tree. But in the attached .cal file, for example, I could not understand why you are assigning very different ages to the same nodes, such as:

-t T1 T7 30.2357

-t T1 T7 73.5762

-t T1 T7 15.835

Cheers

Luis

[Tracy Heath]

In that example, the calibration file assigns multiple fossils to the same node. Each line in the file represents a fossil specimen, not just a "calibration time" as used in conventional methods. This file uses a stochastic branching model that accounts for fossilization as described in this paper: http://arxiv.org/abs/1310.2968. Note that this approach is only appropriate for use with fossil calibration ages. 

Cheers!

Tracy

--
You received this message because you are subscribed to the Google Groups "dppdiv-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dppdiv-users...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[alexandre pedro]

Dear Tracy,

I was revisiting this topic on calibration formats, and I'm having some difficulty to understand the "mean" parameter of the Exponential distribution.

How exactly do I estimate the mean value? Is it equal to the 95% credibility tail that I want to use as a prior (i.e., in your example above, the minimum date would be at 100.2 and the 95% of CI would end at 140 of the exponential tail)? I'm sorry the the naive and rather silly question, but I really tried to understand the exponential function and relate this with the example you provided, but I can't figure that out. What am I missing?

Thank you so much for your time!

Best,

Alex

[Tracy Heath]

Hi Alex,

In that example, when using an exponential distribution as a prior on a calibrated node, the mean parameter is the mean age of the node. Here the expectation of the distribution would be 140 and the 95% highest prior density would be 219.4301. So, based on your prior knowledge of the age of the calibrated nose, you then specify that as the mean for the exponential prior.

Cheers,

Tracy

[Liang LÜ]

Hi Tracy,

I am using pll-dppdiv. But I would like to use fixed ages on some nodes (for example 5 nodes), one for each node, in a simulation study, and I want a normal prior distribution (not expo or uniform). How can I set in the command line and prepare the calibratin file?

Thanks for your time, and best wishes,
Liang Lü

[Tracy Heath]

Hi Liang,

To fix a node to a given value, you just have to set that node as a uniform calibration in the calibration file, with the min and max values both specified as the fixed value. Then, these nodes will be initialized with that time and not updated during the MCMC.

Normal distributions are not implemented for calibration priors. Some of the necessary functions are there, so if you have experience with C++, it shouldn't be difficult to add.

Cheers,

Tracy

[Liang LÜ]

Dear Tracy and all,

I am a little puzzled about the estimattion of dppdiv. The node ages estimated by dppdiv (even under strict or uncorrelated clock models) are mostly overestimated and even two times more than the common sense. So is it necessary to restrict the root age?

Thanks and best wishes,
Liang

[Tracy Heath]

Hi Liang,

This depends very much on your application of calibration priors or fossil taxa (if you used the FBD model). One thing that would make the estimation overly large is if you have incorrectly specified the node on which the calibration should act. Since the DNA data are not informative for the absolute time, it is likely due to how you specified the tree prior or calibrations. I suggest running the analysis under the prior to get a good sense of how these parts of the model are behaving.

Cheers,

Tracy