prep_num_bin: prep_num_bin
In gloverd2/codeBase: Contains standard functions for Admiral Advance Analytics

Description Usage Arguments Value Examples

Function which bins numeric vector into bins of equal weight (or a smooth function on weight).

prep_num_bin(
  var_to_band,
  n_bins = 10,
  weight = rep(1, length(var_to_band)),
  method = "even_weight",
  use_labels = FALSE,
  mean = 0.5,
  sd = 0.3
)

`var_to_band`	Vector[numeric] - vector of values to be binned. Should be true numeric however can also be categorical with fewer levels than `n_bins`
`n_bins`	numeric - Number of bins to split exposure into
`weight`	numeric - vector weight for observations
`method`	string - One of `c("even_weight", "gaussian_weight")`. If `"even_weight"` is used the weight (exposure) in each bucket is even If `"gaussian_weight"` is used the weight look like a normal distribution centered on `mean` with width `sd`. The effect of this is to give thinner tails
`use_labels`	logical - should the bins be numbered (`FALSE`) or human readable label such as [1 - 7] (`TRUE`)
`mean`	numeric - Only used when `method`=`"gaussian_weight"` - value between 0 - 1 for if the bucket with maximum weight is at min or max prediction. Default is 0.5 (middle bucket)
`sd`	numeric - Only used when `method`=`"gaussian_weight"` - value between 0.1 - 1 for how thin the distribution tails should be. Low numbers give thin tails

a list with the following named entries: bins: Either a numeric vector or a factor (depending on use_labels) which gives the bin. The vector is of length length(var_to_band) and has n_bins unique values labels: Human readable labels of where binning has occured cut_values: The numbers the cuts are made at such that they can be reproduced using cut fucntion vector with values 1 to n_bins indicating bin of var_to_band. (1 is low var_to_band and n_bins is high var_to_band)

out1 <- prep_num_bin(var_to_band=1:20, n_bins=5)
out2 <-prep_num_bin(var_to_band=1:20, n_bins=5, method="gaussian_weight")
out3 <-prep_num_bin(var_to_band=1:20, n_bins=5, method="gaussian_weight", use_labels=TRUE)

#To reproduce using cuts
out1$bins
cut(1:20, breaks=out1$cut_values, labels=FALSE)