In pratikunterwegs/klepto-move-evol: R code for the Kleptomove Simulation

Get relative preferences

These are the relative preferences that determine movement strategies.

# load libs
library(data.table)
library(stringr)

# load functions defined locally in R/
devtools::load_all()

Prepare data paths.

# read paths
paths <- list.dirs("data/", recursive = F)
paths <- paths[grep("rep_", paths)]

# make df
data = CJ(
  path = paths
)

# get replicate
data[, rep := as.numeric(
  stringr::str_extract_all(path, "rep_(\\d{3})") |> 
    stringr::str_extract_all("\\d{3}")
)]

# get sim type
data[, sim_type := stringr::str_extract(
  path, "obligate|facultative|foragers|random"
 ) 
]

data[, regrowth := as.numeric(str_extract(path, "(0\\.\\d+)"))]

# filter for focus regrowth
data = data[regrowth == 0.01,]

Relative preference for movement cues

Read scaled movement weights from 2500 individuals in each simulation from generations 1, 10, 30, 100, and 950.

# make a copy
data_rel_pref = copy(data)

# get relative preferences using the function get_scaled_move_prefs
data_rel_pref[, rel_pref := lapply(
  X = path, 

  # apply this function over all the paths
  FUN = get_scaled_move_prefs, 

  # these are arguments to get_scaled_move_prefs
  gens = c(1, 10, 30, 100, 300, 950), 
  n_agents = 2500
)]

# unlist (un-compress) the data to make an ordinary data.frame
data_rel_pref = data_rel_pref[, unlist(rel_pref, recursive = F),
  by = list(sim_type, rep, regrowth)
]

# assign strategy
data_rel_pref[, comp_strat := fifelse(
  sim_type == "obligate", comp_strat, "consumer"
  )
]

fwrite(data_rel_pref, file = "data/results/data_rel_pref.csv")

Evolution of handler tracking

Get the preference for moving towards handlers, i.e., successful foragers, sub-divided by competitive strategy. This is done by processing each simulation replicate using get_pref_handler_by_strat, which is defined and documented in R/fun_syndrome.R

The naming refers to the idea that certain behaviours may be correlated, forming a 'syndrome'.

# run the function over each simulation replicate
syndrome_data <- lapply(
  X = data$path, 

  # apply this function, based around get_pref_handler_by_strat
  FUN = function(path) {
    get_pref_handler_by_strat(
      data_folder = path,
      generations = unique(
        seq(1, 999, 2)
      ),
      weight_comp_strat = 5,
      weight_of_interest = c(
        handler_pref = 3
      ),
      handler_pref_by_strategy = TRUE
  )
})

# make a copy
data_si_strategy = copy(data)

# add to data
data_si_strategy$syndrome_data <- syndrome_data

# unlist
data_si_strategy <- data_si_strategy[, unlist(syndrome_data, recursive = FALSE),
  by = c("sim_type", "rep", "regrowth")
]

# combine proportions and counts for all but the scenario 2 obligate case
data_si_strategy[, comp_strat := fifelse(
  sim_type == "obligate", comp_strat, "consumer"
)]

# sum the number of handler trackers by competitive strategy and
# handler tracking preference (avoid = FALSE, prefer = TRUE)
data_si_strategy = data_si_strategy[, list(
  n_handler_pref = sum(N)
),
by = c(
  "sim_type", "rep", "regrowth", 
  "comp_strat", "handler_pref", "gen"
)
]

# convert to proportions
data_si_strategy[,prop_handler_pref := n_handler_pref / sum(n_handler_pref),
                 by = c(
                   "sim_type", "rep", "regrowth", "gen", "comp_strat"
                 )]

# save data
fwrite(data_si_strategy, file = "data/results/data_handler_tracking_by_strategy.csv")

Relative preferences for competitive strategies

Determine how individuals in scenario 3, who can choose their competition strategy, weigh the presence of other individuals --- successful handler and unsuccessful non-handlers --- when deciding whether to steal or forage.

# make a copy of the data
data_comp_pref = copy(data)

# select for scenario 3 individuals
data_comp_pref = data_comp_pref[sim_type == "facultative",]

data_comp_pref[, comp_pref := lapply(
  X = path, 

  # apply the function get_scaled_move_prefs, which also works for
  # competition strategy weights
  FUN = get_scaled_move_prefs, 

  # arguments to the function, which weights should be scaled
  weights = c(5,6,7,8),
  gens = c(1, 10, 30, 100, 300, 950), 
  n_agents = 2500
)]

# uncompress the data into a regular data frame
data_comp_pref = data_comp_pref[, unlist(comp_pref, recursive = F),
  by = list(sim_type, rep, regrowth)
]

# save to file
fwrite(data_comp_pref, file = "data/results/data_comp_pref.csv")

Calculate competition choices in Scenario 3

Calculate how many individuals would choose to steal or forage given different combinations of food items and handlers, or successful foragers. An individual is a forager if $(w_HH + w_FF) \geq w_0$, assuming there are no non-handlers. We make this assumption to better visualise the outcome on a 2D surface with $F$ and $H$ as the axes.

# read in saved data
data_comp_pref = fread("data/results/data_comp_pref.csv")
# data_comp_pref = data_comp_pref[gen == 950,]

# choice under different number of prey and handlers
data_comp_expand = tidyr::crossing(
  data_comp_pref,
  food = seq(0, 5, 1),
  handlers = seq(0, 5, 1)
)
setDT(data_comp_expand)

# calculate decision; wF = wt_8, and wH = wt_7, w_0 = wt_5
data_comp_expand[, choice := (wt_7 * handlers) + (wt_8 * food) >= wt_5]
data_comp_expand[, choice := fifelse(choice, "forage", "steal")]

# summarise decision: how many individuals would steal or forage
# under different ecological circumstances
data_comp_prop = data_comp_expand[, .N, 
  by = c("gen", "food", "handlers", "choice", "rep")
]

# spread the data into a wide format
data_comp_prop = dcast(
  data_comp_prop,
  gen + food + handlers + rep ~ choice,
  value.var = "N"
)

# save data
fwrite(data_comp_prop, "data/results/data_sc3_comp_choice.csv")