Interpreting PACA & Phylogenetic Signal

[ani.hri...@gmail.com]

Hi everyone,

I'm looking at brain shape variation in the weasel family. I'm trying to interpret the phylogenetic signal and PACA results, and I'd appreciate your input to ensure I'm not way off base here. Data are below as well as PCA, PACA, and k.by.p plots. The first two plots are colored by subfamily. (Side note: Please pardon the coloring, I slapped it on there since I've been unsuccessful in getting R to color my points correctly. I will likely post about that at a later date.) 

Kmult (brain shape) = 0.388

lambda (brain shape) = 0

Kmult Comp1 PACA = 1.15

My main takeaways are as follows. Please let me know if I'm misinterpreting anything.

• The low/moderate Kmult value and lambda = 0 suggest that brain shapes are less similar to one another than expected under Brownian motion. However, phylogenetic structuring of shape is evident in the PCA. Shapes very much parse out by subfamily, so a low phylogenetic signal seems counterintuitive.
  
• Based on this quote from Collyer and Adams, 2020 with my #s pasted in, “The K on the first component (K1) is quite large [~1.15 here], suggesting that Kmult = [0.388] is not a weak phylogenetic signal but rather that the signal was concentrated in some but not all variables of the original data,” this suggests that phylogenetic signal is stronger in some clades than others, right? If correct, I'm trying to reconcile this with evidence of adaptive radiation in subfamilies with a PC1/C1 <0.05 (the left two thirds of the graphs) where I would expect species to be less similar to their close relatives, i.e., have low phylogenetic signal. This bullet point is where I'm having the most trouble reconciling the phylogenetic signal, PCA, and PACA results. 
• When compared to the PCA, the PACA shows the same subfamily phylogenetic structuring along component 1, meaning that phylogeny influences brain shape variation. The PACA also indicates a decrease in variation within subfamilies along component 2, suggesting that any additional variation we see in the PCA is likely due to ecology. 

Thank you to anyone willing to serve as a sounding board. I really do appreciate it.

Sincerely,

Ani

Ani Hristova

PhD Candidate

Department of Integrative Biology

Ecology, Evolution, and Behavior Graduate Program

Michigan State University

[Adams, Dean [EEOB]]

Ani,

I agree with your assessment that the phylogenetic signal in your data is likely concentrated in a few dimensions, and with your assessment of possible ecological information along certain components.

You might also check out our recent paper with Philipp Mitteroecker in Systematic Biology (Mitteroecker, Collyer, and Adams 2025: see my web page), where we devise a means of evaluating phylogenetic signal across dimensions of morphospace. Given the patterns you found with PACA, this will likely show strong and concentrated phylogenetic signal in your data. This is a useful tool to accompany the PACA approach IMO.

The method is available in the latest version of geomorph using the 'physignal.eigen' function.

Best, 

Dean

-- 
Dr. Dean C. Adams
Distinguished Professor
Department of Ecology, Evolution, and Organismal Biology
Iowa State University
www.public.iastate.edu/~dcadams/
phone: 515-294-3834

---

From: geomorph-...@googlegroups.com <geomorph-...@googlegroups.com> on behalf of ani.hri...@gmail.com <ani.hri...@gmail.com>
Sent: Thursday, May 15, 2025 1:23 AM
To: geomorph R package <geomorph-...@googlegroups.com>
Subject: [geomorph-r-package] Interpreting PACA & Phylogenetic Signal

 

--
You received this message because you are subscribed to the Google Groups "geomorph R package" group.
To unsubscribe from this group and stop receiving emails from it, send an email to geomorph-r-pack...@googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/geomorph-r-package/2955f1a4-6a1e-4fc9-96f6-8577d7c0d562n%40googlegroups.com.

[ani.hri...@gmail.com]

Hi Dean & group members,

Thank you for your response; I'm grateful for your insight. I followed your suggestion and evaluated the degree of phylogenetic signal using the physignal.eigen function. It appears that phylogenetic signal is low in my dataset (plot below); however, I'm getting different Kmult values between the physignal and physignal.eigen functions (code & output below). Should this be happening? Or might you have some ideas about why the numbers are calculated differently? I'd like to report the appropriate number for publication. Thanks for your time, and I look forward to hearing from you.

Sincerely,

Ani

---------------------------

Using physignal

summary(physignal(species.dataframe.matched$shape, must.tree, iter = 10000, seed = NULL, print.progress = TRUE))
 
Call:
physignal(A = species.dataframe.matched$shape, phy = must.tree,  
    iter = 10000, seed = NULL, print.progress = TRUE)

Observed Phylogenetic Signal (K): 0.38804
P-value: 1e-04
Based on 10001 random permutations

Using physignal.z (does not provide Kmult)
summary(physignal.z(species.dataframe.matched$shape, must.tree, lambda = "burn", iter = 10000, seed = NULL, print.progress = TRUE))

Call:
physignal.z(A = species.dataframe.matched$shape, phy = must.tree,  
    lambda = "burn", iter = 10000, seed = NULL, print.progress = TRUE)

Evaluation of phylogenetic signal effect size
Optimization method: burn
Optimization performed in 19 data dimensions.
The scaling parameter, lambda, was optimized at 0.
This means that the log-likelihood was invariant across permutations
and there is no phylogenetic signal in the data.

Using physignal.eigen

> physignal.species.shape.evs = physignal.eigen(Y = species.dataframe.matched$shape, phy = must.tree)
> summary(physignal.species.shape.evs)

Observed Phylogenetic Signal (traceK): 7.897
Observed Effect Size (Z-traceK): -0.1272
P-value (traceK): 0.552

Observed Phylogenetic Signal (detK): 3.293e-14
Observed Effect Size (Z-detK): 0.0554
P-value (detK): 0.48

Observed Phylogenetic Signal (Kmult): 0.05657
Observed Effect Size (Z-Kmult): 5.0767
P-value (Kmult): 0.001

Based on 1000 random permutations

[Mike Collyer]

Ani,

What happens if you change the the argument to Blomberg = TRUE, in physignal.eigen?

Mike

<3abc247e-cef5-4a61-abc8-9ba556a42ac1.png>

--
You received this message because you are subscribed to the Google Groups "geomorph R package" group.
To unsubscribe from this group and stop receiving emails from it, send an email to geomorph-r-pack...@googlegroups.com.

To view this discussion, visit https://groups.google.com/d/msgid/geomorph-r-package/daf5abc3-d18c-4494-ab2d-e819305bce31n%40googlegroups.com.
<3abc247e-cef5-4a61-abc8-9ba556a42ac1.png>

[ani.hri...@gmail.com]

Dear Mike,

You're absolutely right, that made the numbers match (and upon re-reading through your 2025 paper, I understand why!). I tried to post the updated output, but my message kept getting deleted, so I'll keep it short. Thank you very much, I appreciate your help!

Sincerely,

Ani

[ani.hri...@gmail.com]

Dear Mike, you're absolutely right! The numbers align now. Thanks very much! 

All the best,

Ani

> physignal.species.shape.evs = physignal.eigen(Y = species.dataframe.matched$shape, phy = must.tree, Blomberg = TRUE)
> summary(physignal.species.shape.evs)

Observed Phylogenetic Signal (traceK): 11.71
Observed Effect Size (Z-traceK): 0.1806
P-value (traceK): 0.47

Observed Phylogenetic Signal (detK): 1.223e-13
Observed Effect Size (Z-detK): 0.0247
P-value (detK): 0.49

Observed Phylogenetic Signal (Kmult): 0.388
Observed Effect Size (Z-Kmult): 4.9061

P-value (Kmult): 0.001

Based on 1000 random permutations

On Wednesday, July 16, 2025 at 1:53:19 PM UTC-4 mlco...@gmail.com wrote: