Generate Comprehensive Diagnostics for B Estimation

Source: R/get_B_diagnostics.R

Generates a comprehensive diagnostic report for B estimation results, including model fit statistics, outlier detection, method comparison (if applicable), and quality control flags.

get_B_diagnostics(
  B_result,
  show_plots = TRUE,
  outlier_threshold = 4,
  verbose = 1
)

Arguments

B_result

Object of class "glow_B_estimate" (from get_B with return_full = TRUE) or "glow_B_model" (from train_B_model). When a glow_B_model is provided, prediction-based diagnostics (method comparison) are skipped.

show_plots

Logical. If TRUE (default), displays diagnostic plots using plot.glow_B_estimate or plot.glow_B_model

outlier_threshold

Numeric. Cook's distance threshold multiplier for outlier detection. Outliers defined as Cook's D > outlier_threshold/n where n is sample size (default = 4, following standard statistical convention)

verbose

Integer. Verbosity level: 0=silent, 1=summary (default), 2=detailed output

Value

A list of class "glow_B_diagnostics" containing:

  • model_fit: List with R^2, adj_R2, formula for each method

  • residual_stats: Summary of residuals (mean, SD, range, normality test)

  • outliers: Indices and details of potential outliers

  • method_comparison: Comparison metrics if both methods used

  • qc_flags: Quality control warnings and recommendations

  • summary_text: Human-readable summary

Details

This function provides a comprehensive assessment of B estimation quality:

Model Fit Statistics:

  • R^2: Proportion of variance explained (higher is better, >0.5 preferred)

  • Adjusted R^2: R^2 adjusted for number of predictors

  • Formula: The selected model formula

Outlier Detection: Uses Cook's distance to identify influential observations that may affect model fit. Threshold is typically 4/n where n is sample size.

Method Comparison (if both methods used):

  • Correlation between methods

  • RMSE (root mean squared error)

  • Percent difference statistics

QC Flags: Automatic warnings for:

  • Low R^2 (<0.3)

  • Non-normal residuals

  • High number of outliers (>10\

  • Large method disagreement (if both methods used)

Computational Complexity

O(n) where n is the number of training variants

See also

get_B for B estimation train_B_model for training B estimation models plot.glow_B_estimate for visualization compare_B_models for comparing all candidate models

Examples

if (FALSE) { # \dontrun{
# Run get_B with full output
result <- get_B(
  training_trait = "binary",
  training_MAF = training_MAF,
  training_BETA = training_BETA,
  training_P = training_P,
  training_N = training_N,
  target_trait = "binary",
  target_MAF = target_MAF,
  target_case_prop = 0.1,
  method = "both",
  return_full = TRUE
)

# Generate diagnostics
diag <- get_B_diagnostics(result, show_plots = TRUE)

# View summary
print(diag)

# Access specific components
diag$model_fit$beta$R2
diag$outliers$indices
diag$qc_flags
} # }