-
Notifications
You must be signed in to change notification settings - Fork 0
/
fisheries.Rmd
192 lines (152 loc) · 8.32 KB
/
fisheries.Rmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
---
title: "Phylogenetic comparative methods in fisheries science"
author: "James T. Thorson"
output: rmarkdown::html_vignette
#output: rmarkdown::pdf_document
vignette: >
%\VignetteIndexEntry{Phylogenetic comparative methods in fisheries science}
%\VignetteEngine{knitr::rmarkdown}
%\VignetteEncoding{UTF-8}
%\VignetteDepends{phylopath,ape,fishtree,phylosignal}
---
```{r, include = FALSE}
have_packages = all(sapply( c("phylopath","ape","fishtree","phylosignal"), FUN=requireNamespace))
knitr::opts_chunk$set(
collapse = TRUE,
comment = "#>"
)
# To isntall
# devtools::install_local( R'(C:\Users\James.Thorson\Desktop\Git\phylosem)', force=TRUE )
# To test build:
# setwd( R'(C:\Users\James.Thorson\Desktop\Git\phylosem)' ); devtools::build_rmd("vignettes/fisheries.Rmd")
# To build PDF:
# library(rmarkdown); setwd( R'(C:\Users\James.Thorson\Desktop\Git\phylosem\vignettes)' ); render( "fisheries.Rmd", pdf_document())
```
Phylogenetic comparative methods (PCM) are rarely in fisheries science, perhaps due to a lack of familiarity with methods and software among fisheries scientists. This underuse is surprising, given that fisheries science and management strongly depends on foundational results from comparative methods regarding fish productivity and life-history parameters, e.g., for defining proxies for biological reference points and informing hard-to-estimate demographic rates such as stock-recruit and natural mortality parameters.
```{r package_warning, include=!have_packages}
message("Must install phylopath, ape, fishtree, phylosignal")
```
## Preparing the Then et al. mortality database
To demonstrate the potential role for PCM in fisheries science, we re-analyze a foundational dataset compiled by [Then et al.](https://www.vims.edu/research/departments/fisheries/programs/mort_db/). We specifically download the file "Mlifehist_ver1.0.csv" and then include a copy as a data object in package _phylosem_ to simplify the following demonstration. This demonstration then provides basic syntax and output from PCM, showing the relationship between natural mortality rate, longevity, and growth parameters.
To do so, we first load a phylogeny for fishes using package _fishtree_, which includes several versions of a phylogeny for fishes develoepd by [Rabosky et al.](https://doi.org/10.1038/s41586-018-0273-1). We then associate all trait data with a tip label from that phylogeny, and provide convenient names for the modeled variables.
```{r, echo=TRUE, results='hide', message=FALSE, fig.width=6, fig.height=6}
# Load packages
library(phylosem)
library(fishtree)
# Download tree
out = fishtree_complete_phylogeny()
tree = out[[1]]
# Load data object
data( Mlifehist_ver1_0 )
Data = Mlifehist_ver1_0
# Reformat to match tree$tip.label
Data$Genus_species = factor( paste0(Data$Genus, "_", Data$Species) )
# Drop duplicates ... not dealing with variation among stocks within species
Data = Data[match(unique(Data$Genus_species),Data$Genus_species), ]
# log-transform to simplify later syuntax
Data = cbind( Data, "logM" = log(Data[,'M']),
"logK" = log(Data[,'K']),
"logtmax" = log(Data[,'tmax']),
"logLinf" = log(Data[,'Linf']) )
# Identify species in both datasets
species_to_use = intersect( tree$tip.label, Data$Genus_species )
species_to_drop = setdiff( Data$Genus_species, tree$tip.label )
# Drop tips not present in trait-data
# Not strictly necessary, but helpful to simplify later plots
tree = ape::keep.tip( tree, tip=species_to_use )
# Drop trait-data not in phylogeny
# Necessary to define correlation among data
rows_to_use = which( Data$Genus_species %in% species_to_use )
Data = Data[rows_to_use,]
# Only include modeled variables in trait-data passed to phylosem
rownames(Data) = Data$Genus_species
Data = Data[,c('logM','logK','logtmax','logLinf')]
```
## Fitting and selecting among phylogenetic structural equation models
We then define a path diagram specifying a set of linkages among variables. In the following, we use a path diagram that ensures that mortality rate is statistically independent of growth, conditional upon a measurement for longevity. This specification ensures that, if longevity is available, then it is the sole information used to predict mortality rate. However, if longevity is not available, the model reverts to predicting mortality from growth parameters.
We then fit this model using phylogenetic structural equation models. We specifically apply a grid-search across the eight models formed by any combination of modeled transformations of the phylogenetic tree. We then use marginal AIC to select a model, and list estimated path coefficients.
```{r, echo=TRUE, results='hide', message=FALSE, fig.width=6, fig.height=6}
# Specify SEM structure
sem_structure = "
logK -> logtmax, b1
logLinf -> logtmax, b2
logtmax -> logM, a
"
# Grid-search model selection using AIC for transformations
Grid = expand.grid( "OU" = c(FALSE,TRUE),
"lambda" = c(FALSE,TRUE),
"kappa" = c(FALSE,TRUE) )
psem_grid = NULL
for( i in 1:nrow(Grid)){
psem_grid[[i]] = phylosem( data=Data,
tree = tree,
sem = sem_structure,
estimate_ou = Grid[i,'OU'],
estimate_lambda = Grid[i,'lambda'],
estimate_kappa = Grid[i,'kappa'],
control = phylosem_control(quiet = TRUE) )
}
# Extract AIC for each model and rank-order by parsimony
Grid$AIC = sapply( psem_grid, \(m) AIC(m) )
Grid = Grid[order(Grid$AIC,decreasing=FALSE),]
# Select model with lowest AIC
psem_best = psem_grid[[as.numeric(rownames(Grid[1,]))]]
```
```{r, echo=FALSE, message=FALSE, fig.width=6, fig.height=6}
knitr::kable(Grid, digits=3, row.names=FALSE)
knitr::kable(summary(psem_best)$coefficients, digits=3)
```
## Visualizing output
Finally, we can convert output to formats from other packages, and use existing and third-party PCM packages to plot, query, and post-process output.
```{r, echo=TRUE, message=FALSE, fig.width=4, fig.height=4}
# Plot path diagram
my_fitted_DAG = as_fitted_DAG(psem_best)
plot(my_fitted_DAG, type="color")
# Total, direct, and indirect effects
my_sem = as_sem(psem_best)
effects(my_sem)
```
We also show how to plot output using phylosignal, although it is currently unavailable on CRAN and therefore commented out:
```{r, eval=requireNamespace("phylosignal",quietly=TRUE), echo=TRUE, message=FALSE, fig.width=6, fig.height=8}
# Load for plotting,
# https://r-pkgs.org/vignettes.html#sec-vignettes-eval-option
library(phylosignal)
# Plot using phylobase
my_phylo4d = as_phylo4d( psem_best )
barplot(my_phylo4d)
```
## Sensitivity to using taxonomic tree
After analysis, we can also conduct sensivity analyses. Here, we show how to construct a taxonomic tree and use this in place of phylogenetic information. As before, this requires some code to reformat the data and then the statistical analysis is simple to specify. The estimated path coefficients are very similar to estimates when using phyogeny.
```{r, echo=TRUE, results='hide', message=FALSE, fig.width=4, fig.height=4}
library(ape)
Data = Mlifehist_ver1_0
# Make taxonomic factors
Data$Genus_species = factor( paste0(Data$Genus, "_", Data$Species) )
Data$Genus = factor( Data$Genus )
Data$Family = factor( Data$Family )
Data$Order = factor( Data$Order )
# Make taxonomic tree
tree = ape::as.phylo( ~Order/Family/Genus/Genus_species, data=Data, collapse=FALSE)
tree$edge.length = rep(1,nrow(tree$edge))
tree = collapse.singles(tree)
tmp = root(tree, node=ape::Ntip(tree)+1 )
# Drop duplicates ... not dealing with variation among stocks within species
Data = Data[match(unique(Data$Genus_species),Data$Genus_species), ]
# log-transform to simplify later syuntax
Data = cbind( Data, "logM" = log(Data[,'M']),
"logK" = log(Data[,'K']),
"logtmax" = log(Data[,'tmax']),
"logLinf" = log(Data[,'Linf']) )
# Only include modeled variables in trait-data passed to phylosem
rownames(Data) = Data$Genus_species
Data = Data[,c('logM','logK','logtmax','logLinf')]
# Fit model
psem_taxon = phylosem( data=Data,
tree = tree,
sem = sem_structure,
estimate_ou = TRUE,
control = phylosem_control(quiet = TRUE) )
# Plot path diagram
my_fitted_DAG = as_fitted_DAG(psem_taxon)
plot(my_fitted_DAG, type="color")
```