Generate Synthetic Data using Bootstrap with Perturbation

Generate Synthetic Data using Bootstrap with Perturbation

Usage

generate_bootstrap_synthetic(
  data,
  continuous_vars,
  cat_vars,
  n = NULL,
  seed = 123,
  noise_level = 0.1,
  id_var = NULL,
  cat_flip_prob = NULL,
  preserve_bounds = TRUE,
  ordinal_vars = NULL
)

Arguments

data

Original dataset to bootstrap from

continuous_vars

Character vector of continuous variable names

cat_vars

Character vector of categorical variable names

n

Number of synthetic observations to generate (default: same as original)

seed

Random seed for reproducibility

noise_level

Noise level for perturbation (0 to 1, default 0.1)

id_var

Optional name of ID variable to regenerate (will be numbered 1:n)

cat_flip_prob

Probability of flipping categorical values (default: noise_level/2)

preserve_bounds

Logical: should continuous variables stay within original bounds? (default: TRUE)

ordinal_vars

Optional character vector of ordinal categorical variables (these will be perturbed to adjacent values rather than randomly flipped)

Value

A data frame with synthetic data

Examples

# \donttest{
# Example 1: Using with GBSG dataset
synth_gbsg <- generate_bootstrap_synthetic(
  data = survival::gbsg,
  continuous_vars = c("age", "size", "nodes", "pgr", "er", "rfstime"),
  cat_vars = c("meno", "hormon", "status"),
  ordinal_vars = c("grade"),
  id_var = "pid",
  n = 1000,
  seed = 123,
  noise_level = 0.15
)
#> Note: The following variables were not specified as continuous or categorical and will be kept as-is from bootstrap sample:
#>    grade

# Example 2: Using with any dataset
my_data <- data.frame(
  id = 1:100,
  height = rnorm(100, 170, 10),
  weight = rnorm(100, 70, 15),
  age = sample(20:80, 100, replace = TRUE),
  gender = sample(c("M", "F"), 100, replace = TRUE),
  education = sample(1:5, 100, replace = TRUE),
  smoker = sample(0:1, 100, replace = TRUE)
)

synth_data <- generate_bootstrap_synthetic(
  data = my_data,
  continuous_vars = c("height", "weight", "age"),
  cat_vars = c("gender", "smoker"),
  ordinal_vars = c("education"),
  id_var = "id",
  n = 150,
  seed = 456
)
#> Note: The following variables were not specified as continuous or categorical and will be kept as-is from bootstrap sample:
#>    education
# }