Run a pop-gen simulation several times and store the results of each in a new column on a data frame

Question

I have a basic Wright-Fisher simulation for two alleles that works wonderfully, and produces a good looking plot showing the allele fixing or dying out by chance as expected. It exports every generation as calculated into a data frame d, so I have the values at each generation to hand. What I want to do is run the whole thing say 20 times and store each complete simulation in a new column automatically, so I can plot all of them on a ggplot graph with colours and all that good stuff. I'm mostly interested in getting a tidy frame to make good-looking plots for a project rather than breakneck efficiency.

#Wright Fisher model Mk1

#Simulation Parameters
# n = pop.size
# f = frequency of focal allele
# x = number of focal allele, do not set by hand
# y = number of the other allele, do not set by hand
# g = generations desired
n = 200
f = 0.6
x = (n*f)
y = (n-x)
g = 200

#This creates a data frame of the correct size to store each generation

d = data.frame(f = rep(0,g))

#Creates the graph.
plot(1,0, type = "n", xlim = c(1,200), ylim = c(0,n),
     xlab = "Generation", ylab = "Frequency A")

#Creates the population, this model is limited to only two alleles, and can only plot one
alleles<- c(rep("A",x), rep("a",y))

#this is the loop that actually simulates the population
#It has code for plotting each generation on the graph as a point 
#Exports the number of focal allele A to the data frame
for (i in 1:g){ 
  alleles <- sample(alleles, n, replace = TRUE)
points(i, length(alleles[alleles=="A"]), pch = 19, col= "red")
F = sum(alleles == "A")
d[i, ] = c(F)
}

So I want to run that last bit multiple times and store each complete iteration somehow. I know I could loop the function by nesting it, even though this is quick and dirty, but doing this only stores the values of the last iteration of the outer loop.

Answer 1

There's a lot of opportunity for improvement here, but this should get you going. I am only showing five simulations, but you should be able to extend. In essence, place the bulk of your code a function and then you can use either map functions from the purrr package or you could also do something with replicate:

library(tidyverse)

n = 200
f = 0.6
x = (n*f)
y = (n-x)
g = 200

d = data.frame(f = rep(0,g))

run_sim <- function() {
  alleles <- c(rep("A", x), rep("a", y))

  for (i in 1:g) { 
    alleles <- sample(alleles, n, replace = TRUE)
    cnt_A = sum(alleles == "A")
    d[i, ] = c(cnt_A)
  }

  return(d)
}

sims <- paste0("sim_", 1:5)

set.seed(4) # for reproducibility

sims %>%
  map_dfc(~ run_sim()) %>%
  set_names(sims) %>%
  gather(simulation, results) %>%
  group_by(simulation) %>%
  mutate(period = row_number()) %>%
  ggplot(., aes(x = period, y = results, group = simulation, color = simulation)) +
  geom_line()

^{Created on 2019-03-21 by the reprex package (v0.2.1)}

NOTE: You could also add arguments to the run_sim function for say x and y (i.e., run_sim <- function(x, y) { ... }), which would allow you to explore other possibilities.