find_optimal_occupancy_thin: Find optimal parameters to calculate occupancy

In bblonder/hypervolume: High Dimensional Geometry, Set Operations, Projection, and Inference Using Kernel Density Estimation, Support Vector Machines, and Convex Hulls

Find optimal parameters to calculate occupancy

Description

The find_optimal_occupancy_thin() function is used to find the optimal parameters for hypervolume_n_occupancy().

Usage

find_optimal_occupancy_thin(..., 
                            verbose = TRUE, 
                            sequence = seq(0, 1, 0.1), 
                            n = 10, 
                            res_type = "raw")

Arguments

...	Parameters to be used to run hypervolume_n_occupancy().
verbose	Logical value; print diagnostic output if TRUE.
sequence	Quantiles to be tested.
n	Number of seeds to be tested.
res_type	If raw print all the seeds and quantiles tested together with the resulting root mean square error (RMSE). If summary print RMSE mean and standard deviation for each quantile.

Details

The find_optimal_occupancy_thin() function searches for the optimal parameters for running hypervolume_n_occupancy(). It works by testing different quantiles and n seeds for random number generation (the same set of n seeds is tested for each quantile). RMSE is returned as the measure of the goodness of fit and results are ordered by increasing RMSE when res_type = "raw". Quantile equal to 0 correspond to no thin. The obtained parameters can be used to feed arguments quant.thin and seed within the function hypervolume_n_occupancy().

Value

A data.frame.

See Also

hypervolume_n_occupancy

Examples

## Not run: 
data(penguins,package='palmerpenguins')
penguins_no_na = as.data.frame(na.omit(penguins))

# split the dataset on species and sex
penguins_no_na_split = split(penguins_no_na, 
                             paste(penguins_no_na$species, penguins_no_na$sex, sep = "_"))

# calculate the hypervolume for each element of the splitted dataset
hv_list = mapply(function(x, y) 
  hypervolume_gaussian(x[, c("bill_length_mm","bill_depth_mm","flipper_length_mm")],
                       samples.per.point=100, name = y), 
  x = penguins_no_na_split, 
  y = names(penguins_no_na_split))


# transform the list into an HypervolumeList
hv_list = hypervolume_join(hv_list)

# find optimal parameters
opt_par = find_optimal_occupancy_thin(hv_list, 
                                       classification = rep(c("female", "male"), 3),
                                       n = 20)

head(opt_par)

unoptimized_hv_occ = hypervolume_n_occupancy(hv_list, 
                        classification = rep(c("female", "male"), 3))

optimized_hv_occ = hypervolume_n_occupancy(hv_list, 
                        classification = rep(c("female", "male"), 3), 
                        quant.thin = opt_par[1, 2], seed = opt_par[1, 1])


## End(Not run)

bblonder/hypervolume documentation built on Feb. 1, 2024, 8:01 p.m.