Detailed feedback presentation?

[richelbilderbeek]

@JolienGay : if and only if you like, I can give some detailed feedback on the presentation. You probably already have a lot you want to change, let me know when you need my precise eye.

[JolienGay]

It will take me a while to finish the ppt for today (after 5 pm) due to some stuff but I would appreciate feedback ofc. Maybe tomorrow when most is done? I planned on adding a few dias with the gamma stuff and ofc the parameterfile example. But other than that I don't know what you mean with "You probably already have a lot you want to change". The only other thing I thought I needed to do today was the poster.

[richelbilderbeek]

Maybe tomorrow when most is done?

Whenever you need it.

[JolienGay]

Ok thx!

[FransThon]

Jolien, the results are going to look quite different, don't worry about that part too much.

[richelbilderbeek]

Just some notes I scribbled down, I put them here for safekeeping:

• There are some references missing, e.g. where is read that the slowdown effect is caused by incomplete sampling?
• There is a slide in which the slowdown in the LTT is shown (with a.o. the dendroica) and compared to the expected LTT. The left figure (the expected average LTT for a BD sim) has a linear y-scale, where the figure from Etienne&Rosindell uses a logarithmic y-scale.
• The reason one individual per species is sampled is not because BEAST2 cannot support this, but it does assume a monophyletic species tree. Here is an example of a paraphyletic tree (at the top), and two sampled species trees

[JolienGay]

Thanks, I will implement it.

[JolienGay]

I think I sometimes confuse some things during the ppt.. So BEAST2 assumes a monophyletic species tree because of the assumptions of the pure BD model. Because of the instant speciation rate not allowing for multiple individuals for a given species?

[richelbilderbeek]

Correct: if speciation is instantaneous, all trees will monophyletic. One can add multiple individuals per species in BEAST2, but then it will infer one monophyletic species tree (see #11), which may be misinformative.

[JolienGay]

Understood! About the age, this was the total length of the tree?

[richelbilderbeek]

Correct. 15 denotes fifteen million years

[JolienGay]

Ok! Last question about the parameter file and I think I get it then. I have highlighted 7 important parameters in an example file: the 5 PBD parameters, the age and mutation rate. The mutation rate is necessary to generate the alignments to get the data into BEAST. Is this correct?

[richelbilderbeek]

the 5 PBD parameters, the age and mutation rate

There is also DNA sequence/alignment length

The mutation rate is necessary to generate the alignments to get the data into BEAST. Is this correct?

Yes

[JolienGay]

Oh right! I'll highlight that one as well, thanks!

[JolienGay]

About the gamma statistics, we do assume the protracted trees should be stemmy right? So gamma > 0.

I don't expect an answer this late ofc ;) but for safekeepings sake! I might forget tomorrow.

[JolienGay]

Still confused, found nice trees in this article: http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0115132#pone-0115132-g002

S1_Figure and S2Figure: I intend to use these as example in the ppt but they explain tippy and stemmy differently than @Femke did. I don't know if this is because I interpret it wrong or Femke confused the drawings, nevertheless I will use the article in the ppt and call it a day.

[FransThon]

No, gamma < 0.

[FransThon]

Ehm... seems like that other article we saw used the terms the other way around...

[FransThon]

http://www.nature.com/ncomms/2015/151117/ncomms9837/full/ncomms9837.html#f3 =/

So which one is it?

[FransThon]

http://www.biologie.ens.fr/phyloeco/pdf/Manceauetal2015ELE.pdf

So, I think it is stemmy if there are many nodes close to the root after all... but that article you found /does/ use the terms that way!

[JolienGay]

If it is like you say, I don't understand it and can't explain it properly tomorrow in the presentation. Richel will help us out of our misery ;P.

[JolienGay]

But, the printed article we got from Richel (Testing macro-evolutionary models using incomplete molecular phylogenies) seems to aggree with the first article. So does this one: http://onlinelibrary.wiley.com/doi/10.1111/j.1466-8238.2012.00759.x/abstract

Again, could be interpreting it wrong, but it took a long time to figure out and I'm a little tired =P.

[FransThon]

It seems like the terms are used to mean both. That's annoying.

[richelbilderbeek]

AFAIKS, one needs a tree with four taxa to being able to explain the difference:

           +--
  +--------+
  |        +--
--+
  |        +--
  +--------+
           +--

    +---------
  +-+
  | +---------
--+
  | +---------
  +-+
    +---------

I'd bet all of us think the same which of these trees is stemmy and tippy?

[richelbilderbeek]

My collection of quotes:

Davies & Buckley 'Exploring the phylogenetic history of mammal species richness':

We find the phylogeny [...] is imbalanced and ‘stemmy’ (long branches towards the root), [...]. In contrast, the [...] is balanced and ‘tippy’ (long branches towards the tips), more consistent with the slow accumulation of diversity over long times [...].

So, upper tree is stemmy

Asmyhr, Linke, Hose, Nipperess 'Systematic Conservation Planning for Groundwater Ecosystems Using Phylogenetic Diversity':

The 18S tree had higher taxa (e.g., Amphipoda) within the tree forming compact groups with long unbranched stems leading up to them (i.e., stemmy tree), whereas the COI tree had relatively longer inter-nodal distances towards the tips of the phylogeny (i.e., tippy tree) [46].

So, upper tree is stemmy. This indeed leads to compact groups.

From Keil, Storch & Jetz, 'On the decline of biodiversity due to area loss':

We predict that if a tree has ‘tippy’ topology (Fig. 3a, orange), then a species that is randomly selected for extinction will, on average, represent lower proportion of the total branch lengths of the tree, compared with an extinction that occurs in a tree that has ‘stemmy’ or rake-like topology (Fig. 3a, green).

IMHO, this explanation is phrased to vague for me to attach value to it.

[richelbilderbeek]

My feedback, feel free to ignore :-) . I've added checkboxes :sunglasses:

• [x] The Pull Of The Present is something else than what you mention. I suggest not to mention it: The POTP is a simulation artefact, that does not exist in nature
• [x] Mention more: BD is a subset of PBD models in which SCR is infinite. The parameter sets with SCR = 1M are therefore special
• [ ] Be more certain about the conclusion: you may conclusively have found that the gamma statistic is recovered correctly. OTOH, how much information is lost when summarizing a tree to a single value?
• [ ] Be less certain about the conclusion: there are many more summary statistics.
• [x] Mention gamma statistic references
• [ ] Perhaps use the simpler trees I put in the demo_gamma_statistic vignette
• [x] Rainbow-colored graph comparing gamma statistic distributions could be zoomed in more

[JolienGay]

Perhaps use the simpler trees I put in the demo_gamma_statistic vignette: I actually really like this example, and everything fits now. Would you be mad (=P) if I let it as it is?

[richelbilderbeek]

As I said: feel free to ignore and follow your own personal preferences.

[richelbilderbeek]

Presentation has been done. Well done!