Advanced Multiple Imputation & Sensitivity Analysis — advancedimputation • ClinicoPath

Advanced multiple imputation by chained equations (MICE) with comprehensive sensitivity analysis for clinical research applications. Includes nested imputation for multilevel data, missing not at random (MNAR) imputation methods, and comprehensive sensitivity testing for missing data assumptions. Essential for regulatory-compliant clinical research where missing data handling must be thoroughly documented and validated.

Usage

advancedimputation(
  data,
  imputation_vars,
  auxiliary_vars,
  cluster_var,
  id_var,
  n_imputations = 10,
  n_iterations = 10,
  convergence_check = TRUE,
  imputation_method = "pmm",
  categorical_method = "logreg",
  mnar_methods = FALSE,
  mnar_type = "delta_adjustment",
  delta_values = "0",
  sensitivity_analysis = TRUE,
  sensitivity_methods = "all",
  multilevel_imputation = FALSE,
  level1_vars,
  level2_vars,
  min_bucket_size = 5,
  exclude_vars,
  passive_imputation,
  ridge_penalty = 1e-05,
  remove_collinear = TRUE,
  collinearity_threshold = 0.95,
  diagnostic_plots = TRUE,
  imputation_quality = TRUE,
  cross_validation = FALSE,
  ampute_test = FALSE,
  save_imputations = FALSE,
  pool_results = TRUE,
  show_detailed_output = TRUE,
  regulatory_report = FALSE,
  random_seed = 123
)

Arguments

data: the data as a data frame
imputation_vars: Variables with missing values to be imputed
auxiliary_vars: Complete or mostly complete variables to assist imputation
cluster_var: Clustering variable for nested/multilevel imputation (e.g., study site, patient ID)
id_var: Subject identifier for tracking observations across imputations
n_imputations: Number of imputed datasets to create (recommended 5-20 for analysis, 100+ for final results)
n_iterations: Number of MICE iterations per imputation (increase for convergence)
convergence_check: Monitor and assess MICE convergence with diagnostic plots and statistics
imputation_method: Primary method for continuous variables
categorical_method: Method for categorical variables
mnar_methods: Include Missing Not At Random imputation methods
mnar_type: Type of MNAR imputation approach
delta_values: Comma-separated delta values for MNAR sensitivity analysis (e.g., "0, -0.5, -1")
sensitivity_analysis: Perform extensive sensitivity analysis for missing data assumptions
sensitivity_methods: Type of sensitivity analysis to perform
multilevel_imputation: Use multilevel imputation for nested/clustered data
level1_vars: Individual-level variables for multilevel imputation
level2_vars: Cluster-level variables for multilevel imputation
min_bucket_size: Minimum number of donors for PMM (prevents poor matching)
exclude_vars: Variables to exclude from imputation model (but keep in dataset)
passive_imputation: R expressions for passive imputation (e.g., "bmi ~ I(weight/height^2)")
ridge_penalty: Ridge penalty for numerical stability in regression imputation
remove_collinear: Automatically remove highly collinear variables from imputation model
collinearity_threshold: Correlation threshold for removing collinear variables
diagnostic_plots: Generate convergence and diagnostic plots
imputation_quality: Assess quality of imputations vs observed data
cross_validation: Perform cross-validation of imputation methods
ampute_test: Test imputation performance using artificial missingness
save_imputations: Save completed imputed datasets for further analysis
pool_results: Pool results across imputations using Rubin's rules
show_detailed_output: Show detailed imputation summaries and diagnostics
regulatory_report: Generate regulatory-compliant missing data analysis report
random_seed: Set random seed for reproducible results

Value

A results object containing:

`results$instructions`					Instructions for advanced multiple imputation and sensitivity analysis
`results$missing_summary`					Overview of missing data patterns and completeness
`results$convergence_assessment`					MICE convergence diagnostics and statistics
`results$imputation_summary`					Summary of completed imputations across methods
`results$sensitivity_results`					Results across different imputation methods and assumptions
`results$mnar_analysis`					Missing Not At Random sensitivity analysis results
`results$multilevel_results`					Results for nested/multilevel imputation
`results$quality_assessment`					Comprehensive quality evaluation of imputations
`results$cross_validation_results`					Cross-validation performance of imputation methods
`results$amputation_test_results`					Performance assessment using artificial missingness
`results$pooled_estimates`					Results pooled across imputations using Rubin's rules
`results$convergence_plot`					MICE convergence trace plots and diagnostics
`results$distribution_plot`					Comparison of observed and imputed value distributions
`results$pattern_plot`					Visual representation of missing data patterns
`results$sensitivity_plot`					Visualization of sensitivity analysis results
`results$quality_plot`					Quality assessment visualization across variables
`results$clinical_interpretation`					Clinical context and imputation recommendations
`results$regulatory_report`					Comprehensive report for regulatory submissions
`results$methods_documentation`					Detailed documentation of imputation methods and assumptions

Tables can be converted to data frames with asDF or as.data.frame. For example:

results$missing_summary$asDF

as.data.frame(results$missing_summary)

Examples

data('clinical_trial_data')

advancedimputation(
    data = clinical_trial_data,
    imputation_vars = c("primary_endpoint", "biomarker", "age"),
    auxiliary_vars = c("baseline_score", "treatment_group"),
    sensitivity_analysis = TRUE,
    mnar_methods = TRUE
)