interval time outputs

[quattrinia]

Hi,

I am a little confused what interval times this function (rev.process.div.rates) returns. I put in the following interval times in my ER div analysis
interval_times <- v(40, 66, 180, 200, 358, 510, 542, 600)
and set the num_intervals to 9 (Assuming 0-40 is the first interval, etc.)
But after I process the files using rev_out function, the intervals are different
$intervals
[1] 794.44402 706.17246 617.90090 529.62935 441.35779 353.08623 264.81467 176.54312 88.27156 0.00000
I don't understand why these times are different. (The output log files have the correct times) Is this a bug or am I missing something?

[quattrinia]

Ok. I understand that this is breaking up the tree into set number of intervals, but if you set particular intervals in the analysis, as I did above, is there a way that this can be reflected in the plots?

[hoehna]

Hi, currently RevGadgets is breaking the intervals into equal spaced epochs, i.e., each epoch has the same length. We do not (yet) allow you to provide your own intervals. The reason for the current setting was that the interval times might change during the MCMC analysis. Therefore, RevGadgets uses the equal sized intervals for post-processing and plotting purposes. For your example, it could work if you use a large number of intervals in RevGadgets, for example, if you use the tree age as the number of intervals. In that case you will get one interval per year, and the first 40 intervals should all be equal, etc. This might not be the most efficient way, but should work to plot the rates for your given settings. Best, Sebastian

…

On Wed, Jan 8, 2020 at 9:18 PM quattrinia ***@***.***> wrote: Ok. I understand that this is breaking up the tree into set number of intervals, but if you set particular intervals in the analysis, as I did above, is there a way that this can be reflected in the plots? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <revbayes/RevGadgets#4?email_source=notifications&email_token=AB6P5R7YCA4S4EWT5ECJGN3Q4YYJFA5CNFSM4KEN5VHKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEIN22BY#issuecomment-572239111>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB6P5R6NITJLUZVA2VWIG63Q4YYJFANCNFSM4KEN5VHA> .

[quattrinia]

Ok. I understand, but I am getting pretty different plots if I break out the times into different number of intervals.
And the plots are attached, demonstrating how different these look if I just plot with either 9 intervals or say 20 intervals. Advice?

FYI, I compared this model to two others, with settings that included different time intervals and equal spaced intervals, and this model did not perform as well. So, this might all just be irrelevant, but would like to know how to proceed for future....
EBD_9in.pdf
EBD_20in.pdf

[hoehna]

The difference in the results in not surprising. If you choose a low number of intervals, then the intervals are larger and thus average more over the rate variation. This is because we compute the weighted average rate in each interval. The recommendation is that you use as many intervals for plotting as possible. I'm usually using 100 or 1000 intervals. The plotting with many intervals only shows you the variation and uncertainty in your estimated rates. If you choose a too low number, you might accidentally hide or smooth out estimated rate variation. If you are concerned about the variation in estimated rates, then you could try some other prior models (see here: https://www.biorxiv.org/content/10.1101/853960v1.abstract). Best, Sebastian

…

On Mon, Jan 13, 2020 at 10:58 PM quattrinia ***@***.***> wrote: Ok. I understand, but I am getting pretty different plots if I break out the times into different number of intervals. And the plots are attached, demonstrating how different these look if I just plot with either 9 intervals or say 20 intervals. Advice? FYI, I compared this model to two others, with settings that included different time intervals and equal spaced intervals, and this model did not perform as well. So, this might all just be irrelevant, but would like to know how to proceed for future.... EBD_9in.pdf <https://github.com/revbayes/RevGadgets/files/4056013/EBD_9in.pdf> EBD_20in.pdf <https://github.com/revbayes/RevGadgets/files/4056014/EBD_20in.pdf> — You are receiving this because you commented. Reply to this email directly, view it on GitHub <revbayes/RevGadgets#4?email_source=notifications&email_token=AB6P5R7I43OGZTXYG2MAVGLQ5TPY5A5CNFSM4KEN5VHKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEI2OJOA#issuecomment-573891768>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB6P5RYYRMBSJUHJL2TID63Q5TPY5ANCNFSM4KEN5VHA> .

[quattrinia]

Ah, ok. Thanks for the help!