Introduction

We will explore several attributes of the R (R Core Team 2020) package mvSLOUCH (Bartoszek et al. 2012), that allows fitting multivariate models for phylogenetic comparative data with emphasis on those based on an Ornstein-Uhlenbeck process. Versions of mvSLOUCH from 2.0.0 run models at considerably faster speeds through using the computational engine provided by PCMBase (Mitov et al. 2020), so let us start by attaching mvSLOUCH as well as
ggplot2 (Wickham 2016), which have some useful functions for data processing and plotting, respectively (PCMBase with its suggested PCMBaseCpp, if installed, C++ backend will be loaded by mvSLOUCH):

library(ggplot2)
library(ape)
library(mvSLOUCH)

## Loading required package: abind

As a phylogenetic comparative analysis with mvSLOUCH can run a long time, we first load required fragments of the precalculated objects. The unreduced objects are the results of the estimation procedures run here with the set random seed. We preload them so that building the vignette does not take over a day and we store within mvSLOUCH only the necessary parts as the full objects would take up too much space. The complete objects can be found in the KJVJMRKK_mvSLOUCH GitHub repository. However, the readers are encouraged to run the presented code by themselves.

load("./Carnivora_mvSLOUCH_objects.RData")

Using an example on carnivoran locomotion and forelimb morphology, we will explore and compare the basic models of mvSLOUCH, going through some key tasks associated with its inputs (e.g. setting up the data and the phylogeny, specifying selective regimes for adaptive hypotheses) and outputs (e.g. identifying key statistics, optimizing parameter estimates of a preferred model, computing confidence intervals under parametric bootstrap).

The function set.seed() allows specifying a starting point in a sequence of randomly generated numbers so that a user can obtain the same outputs under a given process. For the purposes of the current vignette, if you want to replicate the outputs below (for mvSLOUCH 2.6.2), set up the following seed:

RNGversion(min(as.character(getRversion()),"3.6.1"))
set.seed(5, kind = "Mersenne-Twister", normal.kind = "Inversion")

Data: Locomotion and forelimb morphology in carnivorans

The order Carnivora has colonized a wide variety of habitats. The specific challenges of moving through these habitats are reflected in the diversity of locomotor strategies they exhibit, such as running fast or for long distances (i.e. cursorial locomotion, such as hyenas and wolves), climbing (e.g. from the scansorial raccoon to the fully arboreal kinkajou) swimming (e.g. from the semiaquatic otter to the fully aquatic seal) and digging (i.e. semifossorial locomotion, such as some skunks and badgers). Although some morphological attributes are useful across different locomotor types (e.g. swimmers and diggers are similarly benefited by an increased area of the paws and high force outputs of the limbs), others can be at odds with each other. A good example of the latter involves cursorial and semifossorial carnivorans. At the core of the contrast between these two locomotor types lies a trade-off in limb mechanics, as many adaptations that maximize velocity transmissions are at odds with those that maximize force outputs. Limb bones selected for strength exhibit pronounced crests that increase the area of insertion for locomotor muscles, and limb segments tend to be short given that smaller output levers result in higher force outputs. On the other hand, elongated limb segments result in longer strides and larger output levers that favor runners (because larger output levers increase relative velocity transmissions). Runners also benefit from lighter limbs that maximize the distance gained per force input of each stride i.e. big muscles and conspicuous crests are less advantageous for runners as they are for diggers. The mvSLOUCH package offers analytical tools for evaluating evolutionary hypotheses that both:

1. relate limb morphologies to locomotor types (e.g. runners vs diggers); and
2. test for trade-offs between force outputs and velocity transmissions in limb mechanics (i.e. strength vs speed).

We will explore these ideas using a subset of the dataset collected by Samuels et al. (2013), available at the Dryad Data Repository dx.doi.org/10.5061/dryad.77tm4. We first download the data, remove fossil samples (lacking locomotor ecology data; Urocyon cinereoargenteus is removed separately as the current and fossil samples have the same entry as species name in the data file), species with missing measurements (for radius length, see below) and Lycalopex sp. (as species identification is required for branch length assignation), rename species according to Johnson et al. (2006) and Wilson and Reeder (2005), and match the tip labels in the phylogeny (spaces replaced by underscores), and finally keep only the columns that we need (locomotor habits, humerus length, deltopectoral crest length and radius length of the forelimb; see below).

b_correct_dryad_download<-FALSE
temp <- tempfile()
tryCatch({
download.file("datadryad.org/api/v2/datasets/doi%253A10.5061%252Fdryad.77tm4/download",temp)
b_correct_dryad_download<-TRUE
},error=function(e){cat("Could not download data file from Dryad! No analysis can be done! Vignette not built!")},
warning=function(w){cat("Problem with downloading data file from Dryad! No analysis can be done! Vignette not built!")})

if (b_correct_dryad_download){
    b_correct_dryad_download<-FALSE
    tryCatch({
    dfcarnivores_postcranial <- read.table(unz(temp, "Carnivore Postcranial Morphology Data Samuels et al. 2013.txt"),header=TRUE,sep="\t",stringsAsFactors =FALSE)
    b_correct_dryad_download<-TRUE
    },error=function(e){cat("Corrupt data file from Dryad! No analysis can be done! Vignette not built!")},
    warning=function(w){cat("Problem with accessing data file from Dryad! No analysis can be done! Vignette not built!")})
}

## Error: Could not download data file from Dryad! No analysis can be done! Vignette not built!