Analyze Data with Multiple Imputation

Applies an analysis function to each imputed dataset and pools results using Rubin's rules.

Usage

analyze_with_imputation(imputation_result, analysis_fun, conf_level = 0.95)

Arguments

imputation_result

An ImputationResult object from perform_multiple_imputation()

analysis_fun

A function that takes a completed data frame and returns a list with 'estimate' and 'variance' components

conf_level

Numeric. Confidence level for pooled CI (default: 0.95)

Value

A list containing pooled results from Rubin's rules

Details

The analysis_fun must return a list with at least:

• estimate: The point estimate from the analysis

• variance: The variance of the estimate

For regression models, this typically means extracting the coefficient and its squared standard error.

Examples

if (FALSE) { # \dontrun{
data <- data.frame(
  outcome = c(10, NA, 30, 40, 50),
  predictor = c(1, 2, NA, 4, 5)
)

imp <- perform_multiple_imputation(data, m = 5)

# Analyze with linear regression
result <- analyze_with_imputation(imp, function(d) {
  fit <- lm(outcome ~ predictor, data = d)
  coefs <- summary(fit)$coefficients
  list(
    estimate = coefs["predictor", "Estimate"],
    variance = coefs["predictor", "Std. Error"]^2
  )
})

result$pooled_estimate
result$ci
} # }