Show how to do more general types of selection (balancing, diverging etc.)

[gregorgorjanc]

We usually simulate directional selection.

In some cases, we might want to simulate balancing selection or diverging selection etc. While these types of selection can be done with the selectInd() and some scripting, it would be instructive to show how to use trait = selectionFun(pop) ... to simulate different types of selection - we just need an appropriate selectionFun() ...

[gaynorr]

To expand on your comments for the benefit of others, here is an example of how it can currently be done in AlphaSimR with the abs function as your selection function. This approach only works for a single trait. For multiple traits, you'd have to write your own selection function. The specifics of how to write such a function haven't really been documented. I'm not sure where best to document this.

As an aside, creating a selection function that returns a single value as a function of multiple traits is a great way to incorporate crop modeling approaches and/or simulate complex epistasis.

library(AlphaSimR)

founderPop = runMacs(nInd = 1000, nChr=10, segSites = 1000)

SP = SimParam$
  new(founderPop)$
  addTraitA(nQtlPerChr = 1000)$
  setVarE(h2 = 0.3)

# Perform stabilizing selection, minimize absolute value
pop = newPop(founderPop)
hist( gv(pop) )

gain = numeric(50)
for(gen in 1:50){
  pop = selectCross(pop, nInd = 100, nCrosses = 1000, 
                    trait = abs, selectTop = FALSE)
  
  gain[gen] = mean( abs( gv(pop) ) )
}
hist( gv(pop) )
plot(1:50, gain, type="l", xlab="Generation", ylab="Trait Mean")

# Perform divergent selection, maximize absolute value
# Likely to randomly go to one of the trait extremes
pop = newPop(founderPop)
hist( gv(pop) )

gain = numeric(50)
for(gen in 1:50){
  pop = selectCross(pop, nInd = 100, nCrosses = 1000, 
                    trait = abs)
  
  gain[gen] = mean( abs( gv(pop) ) )
}
hist( gv(pop) )
plot(1:50, gain, type="l", xlab="Generation", ylab="Trait Mean")

[gaynorr]

My divergent selection example isn't the best, because it will likely just be directional selection in a random direction.

[gregorgorjanc]

As to the documenting this behaviour, I think best to add a simple example to selectInd() examples section in the docs and/or link to function like selectionIndex(). I will open a PR for this.

[gregorgorjanc]

PR in #148