NLRX Package: Include boolean parameter as variable-CodePudding

Can I include boolean parameters in the variables for a simdesign? An if so, how would that look like? Right now I only have metric variables like that:

variables = list(
    'navigation-system-rate' = list(min = 0, max = 1, qfun="qunif"),
    "expected-weight" = list(min = 0, max = 1, qfun="qunif")
  )

but I want them to take the value "true" or "false" randomly for a certain parameter. Is that possible somehow?

CodePudding user response：

Yes you can define boolean and categorical variable values, but only for certain simdesigns (simple, distinct, full factorial) as explained here: https://docs.ropensci.org/nlrx/articles/furthernotes.html#variables-and-constants-definitions

For example you can use,

variables = list(
    "bool1" = list(values = c("true", "false")),
    "bool2" = list(values = c("true", "false"))
  )

and then for example apply a simdesign_ff() to run all combinations of your boolean variables.

All other simdesigns expect a numeric range/distribution and will not work with any categorical/boolean values. As a simple workaround you could just apply some transformation to your parameter matrix after it has been generated.

E.g., if you are doing a latin hypercube, you can just define your boolean variable as uniformly distributed variable within the range 0 to 1. Then after applying the simdesign_lhs, just overwrite the nl@simdesign@siminput tibble and update the values of your variable by setting all cells with a value < 0.5 to false, and all other values to true.


# Define variable as numeric in range 0 .. 1
nl@experiment <- experiment(
  # ...
  variables = list("bool1" = list(min = 0, max = 1, qfun="qunif"))
  # ...
)
                         
# apply simdesign_lhs to generate parameter matrix
nl@simdesign <- simdesign_lhs(nl=nl,
                               samples=100,
                               nseeds=3,
                               precision=3)

# apply transformation to boolean (netlogo needs booleans as strings)
nl@simdesign@siminput <- nl@simdesign@siminput %>% 
  dplyr::mutate(bool1 = dplyr::if_else(bool1 < 0.5, "false", "true"))