Demo: Fitting clade-specific and island-wide diversity-dependence models in DAISIE

1. Load, prepare and visualize data

• If you want to start directly with model fitting you can skip this step and go to step 2.

Load the phylogenetic data for Eleutherodacylus frogs from the island of Hispaniola (used in Etienne et al. 2020). This dataset contains times of colonization and branching times for all species of Eleutherodactylus frogs found on the island. The data was extracted from the dated molecular phylogeny of Dugo-Cota et al 2019 (Ecology Letters 22:884–893). The dataset includes 65 extant species, which are the result of five independent colonisation events of the island of Hispaniola.

data(frogs_datatable, package = "DAISIE")

You can load your own data in a table/tibble format, making sure the table headers match the ones in the example.

1.1. Create DAISIE datalist from the input table

In this case we are assuming an island age of 30 million years for Hispaniola and a mainland pool size of 300 species (this is the number of species that may colonize the island).

frogs_datalist <- DAISIE_dataprep(
        datatable = frogs_datatable,
        island_age = 30,
        M = 300)

1.2. View DAISIE datalist

Just type:

frogs_datalist

The DAISIE datalist includes the data from the table frogs in the format of a DAISIE object that can be read in the subsequent maximum likelihood functions (e.g DAISIE_ML, DAISIE_ML_IW ). It includes the age of the island, the number of mainland species that are not currently present on the island and a list with all extant independent colonisation events, including their colonization and branching times.

1.3. Visualise the data

DAISIE_plot_island(frogs_datalist)

This plot shows the different colonization events, their times of colonization and branching times (horizontal ticks). Dashed line shows the age of the island. n=number of species; m= number of species missing.

# DAISIE_plot_age_diversity(frogs_datalist)

2. Fit DAISIE models

If you skipped step 1, first load the Hispaniola Eleutherodactylus datalist.

data(frogs_datalist)

_

We will fit five different DAISIE models to the phylogenetic data contained in frogs_datalist:

• DI - diversity-independent model. Model where K’ (carrying-capacity) is set to Inf.
• CS - clade-specific diversity-dependence model.
• CS_no_ana - clade-specific diversity-dependence model with no anagenesis. Equivalent to the CS model in Etienne et al. 2020.
• IW - island-wide diversity-dependence model.
• IW_no_ana - island-wide diversity-dependence model with no anagenesis. Equivalent to the IW model in Etienne et al. 2020.

We use the DAISIE_ML and DAISIE_ML_IW functions to optimise the likelihood. These are the most important settings to specify in these functions:

• datalist - The name of the DAISIE datalist (in this case frogs_datalist)
• initparsopt - these are the initial values from the parameters for which the likelihood will be optimised. In the examples below we will use parameters similar to the maximum likelihood parameters from the corresponding models in Etienne et al 2020. However, we recommend you try a variety of initial starting parameters to ensure the optimum is found. Note that certain combinations of initial starting values may fail, as parameters must be feasible (e.g. a K’ value lower than the number of species found in the data will not run). In DAISIE the parameters have the following position:
  1. Rate of cladogenesis (per species on the island per time unit)
  2. Rate of extinction (per species on the island per time unit)
  3. Carrying-capacity (K’) - maximum number of species any clade can reach on the island for CS models; maximum number of species on the island across all clades for IW models.
  4. Rate of colonisation (per species on the mainland per time unit)
  5. Rate of anagenesis (per species on the island per time unit)
• idparsopt - The position of the parameters to optimise (e.g. to optimize cladogenesis and anagenesis only c(1,5).
• parsfix - If parameters are being fixed, specify here the value (e.g. if fixing K’ to Inf and fixing anagenesis to 0 - c(Inf,0).
• idparsfix - The position of the paratemers fixed (e.g. to fix K’ and anagenesis c(3,5)).
• ddmodel - Set 0 for diversity-independent models; 11 for diversity-dependent models (IW and CS) where cladogenesis and colonisation decline linearly with diversity. There are other options.
• There are many other options, check ? DAISIE_ML

The IW models (IW and IW_no_ana) use the island-wide version of DAISIE, which is computationally demanding. They require high memory and long run times. We recommend these are run on a cluster. The CS models can be run on a regular computer or laptop.

2.1 Fit DI model

This model contains 4 parameters:
1 - cladogenesis
2 - extinction
4 - colonisation
5 - anagenesis
K’ is fixed to Inf
Set ddmodel=0

DAISIE_ML(
  datalist = frogs_datalist,
  initparsopt = c(0.18,0.03,0.0006,2),
  idparsopt = c(1,2,4,5),
  ddmodel = 0,
  parsfix = Inf,
  idparsfix = 3
)

#>    lambda_c         mu   K        gamma lambda_a    loglik df conv
#> 1 0.1789218 0.02509053 Inf 0.0006259365  1568.88 -209.8634  4    0

2.2 Fit CS model

This model contains 5 parameters:
1 - cladogenesis
2 - extinction
3 - K’
4 - colonisation
5 - anagenesis
Set ddmodel=11

DAISIE_ML(
  datalist = frogs_datalist,
  initparsopt = c(0.44,0.11,36.44,0.0007,2),
  idparsopt = c(1,2,3,4,5),
  ddmodel = 11,
  parsfix = NULL,
  idparsfix = NULL
)

Output not shown here.  

2.3 Fit CS_no_ana model

This model contains 4 parameters:
1 - cladogenesis
2 - extinction
3 - K’
4 - colonisation
Set ddmodel=11

DAISIE_ML(
  datalist = frogs_datalist,
  initparsopt = c(0.44,0.11,36.44,0.0007),
  idparsopt = c(1,2,3,4),
  ddmodel = 11,
  parsfix = 0,
  idparsfix = 5
)

Output not shown here.

2.4 Fit IW model

This model contains 5 parameters:
1 - cladogenesis
2 - extinction
3 - K’
4 - colonisation
5 - anagenesis
Set ddmodel=11

DAISIE_ML_IW(
  datalist = frogs_datalist,
  initparsopt = c(0.41, 0.17, 131.7, 0.0012, 2),
  idparsopt = c(1,2,3,4,5),
  ddmodel = 11,
  parsfix = NULL,
  idparsfix = NULL
)

Output not shown here (recommended to be run on a cluster).

2.5 Fit IW_no_ana model

This model contains 4 parameters:
1 - cladogenesis
2 - extinction
3 - K’
4 - colonisation
Set ddmodel=11

DAISIE_ML_IW(
  datalist = frogs_datalist,
  initparsopt = c(0.40, 0.17, 131.83, 0.0012),
  idparsopt = c(1,2,3,4),
  ddmodel = 11,
  parsfix = 0,
  idparsfix = 5
)

Output not shown here (recommended to be run on a cluster).  

3. Simulate islands under given DAISIE models

set.seed(1)
frog_sims_CS <- DAISIE_sim_constant_rate(
  time=30,
  M=300,
  pars=c(0.44,0.11,36.44,0.0007,0),
  divdepmodel = "CS",
  replicates= 100,
  plot_sims = FALSE)

set.seed(1)
frog_sims_CS <- DAISIE_sim_constant_rate(
  time=30,
  M=300,
  pars=c(0.40,0.17,131.83,0.0012,0),
  divdepmodel = "IW",
  replicates= 100,
  plot_sims = FALSE)

DAISIE_plot_sims(frog_sims_CS)