Skip to contents

Validate a Data Bundle

Source: R/contracts.R
validate_data_bundle.Rd

Validates that a data_bundle conforms to the required structure for use in simulation tasks. The data_bundle is the output of a data_generator function and contains all data-related objects needed for model fitting and metric computation.

Usage

validate_data_bundle(data_bundle)

Arguments

data_bundle

A list containing data and metadata for a simulation task.

Value

The input data_bundle, invisibly, if validation passes.

Details

The data_bundle must have the following structure:

  • train: A data.frame with at least 1 row (required)

  • test: NULL or a data.frame (optional)

  • response: A scalar character naming the response column in train (and test if not NULL)

  • true_params: A named numeric vector where names exactly match vars_of_interest

  • vars_of_interest: A non-empty character vector of unique names

  • references: A named numeric vector where names exactly match vars_of_interest (optional)

  • meta: Optional named list with scalar values only

Validation rules:

  • train must be a data.frame with nrow >= 1

  • test must be NULL or a data.frame

  • response must be a scalar character present in train (and test if not NULL)

  • true_params must be a named numeric vector

  • vars_of_interest must be a non-empty unique character vector

  • setequal(names(true_params), vars_of_interest) must be TRUE

  • If provided, setequal(names(references), vars_of_interest) must be TRUE

  • No duplicate names in any named vector/list

  • meta must be a named list with scalar values only (if present)

Errors

Throws a bayesim_data_error condition if validation fails.

Examples

# Valid data bundle
data_bundle <- list(
  train = data.frame(x = 1:10, y = rnorm(10)),
  test = data.frame(x = 11:15, y = rnorm(5)),
  response = "y",
  true_params = c(beta = 1.5, sigma = 0.5),
  vars_of_interest = c("beta", "sigma"),
  references = c(beta = 0, sigma = 1)
)
validate_data_bundle(data_bundle)