Generate a matrix containing weight combinations — generate

generate_weights() generates a matrix containing weight combinations for a set of variables such that each set of weights sums to 1. This can be supplied to calc_all_lscps to calculate fitness landscapes corresponding to a variety of possible sets of weights for weighting functional characteristics. The weights are generated by partitioning a weight of 1 across however many variables are requested in all possible ways.

Usage

generate_weights(step, n, data = NULL, nvar = NULL,
                 varnames = NULL, verbose = TRUE)

Arguments

step: numeric. The step size between weight partitions. Only one of step and n can be specified.
n: numeric. The number of weight partitions between 0-1. Only one of step and n can be specified.
data: an optional fnc_df (the output of as_fnc_df) or kriged_surfaces (the output krige_surf) object. The number of variabes and their names will be extracted from the data as the functional characteristics present in them.
nvar: the number of variables across which to allocate the weights. Ignored if data is not NULL. If nvar = NULL and varnames is supplied, the length of varnames will be used for nvar.
varnames: the names of the variables across which to allocate the weights. Ignored if data is not NULL. If varnames = NULL and nvar is supplied, the sequence from 1 to nar will be used for varnames.
verbose: whether to display a message noting the number of sets of weights created.

Details

generate_weights() works by fining all possible allocations of n objects into nvar bins. When step is supplied, n is computed as round(1/step), so the resulting weight partitions may not be exactly equal to step when its inverse is not an integer. The larger n is (or the smaller step) is, the more possible allocations will be produced (i.e., and the resulting object will have more rows). The output of generate_weights() can quickly become very large with increasing number of variables, and will make subsequent analyses slow. It is recommended to start with a large step size, or small n, and increment up.

Value

A grid_weights object, which is a matrix with a row for each each set of weights and a column for each variable over which the weights are allocated. The weights in each row will sum to 1.

Examples

# Allocating 10 partitions of .1 across 3 variables
wmat <- generate_weights(n = 10, nvar = 3)
#> 66 rows generated
head(wmat)
#>        1   2 3
#> [1,] 1.0 0.0 0
#> [2,] 0.9 0.1 0
#> [3,] 0.8 0.2 0
#> [4,] 0.7 0.3 0
#> [5,] 0.6 0.4 0
#> [6,] 0.5 0.5 0

# Allocating 5 partitions of .2 across the 4 functional
# characteristics in the warps dataset
data("warps")

warps_fnc <- as_fnc_df(warps)
wmat <- generate_weights(n = 5, data = warps_fnc)
#> 56 rows generated
head(wmat)
#>      hydro curve mech fea
#> [1,]   1.0   0.0    0   0
#> [2,]   0.8   0.2    0   0
#> [3,]   0.6   0.4    0   0
#> [4,]   0.4   0.6    0   0
#> [5,]   0.2   0.8    0   0
#> [6,]   0.0   1.0    0   0

# Using 'step' for the same result:
wmat <- generate_weights(step = .2, data = warps_fnc)
#> 56 rows generated
head(wmat)
#>      hydro curve mech fea
#> [1,]   1.0   0.0    0   0
#> [2,]   0.8   0.2    0   0
#> [3,]   0.6   0.4    0   0
#> [4,]   0.4   0.6    0   0
#> [5,]   0.2   0.8    0   0
#> [6,]   0.0   1.0    0   0