Package {accmv}

Inference with Available Complete-Case Missing Values

Description

Implements inverse-probability-weighted, regression-adjustment, and multiply-robust estimators under the available complete-case missing value assumption of Cheng, Chen, Smith, and Zhao (2022).

References

Cheng G, Chen Y-C, Smith MA, Zhao Y-Q (2022). Handling Nonmonotone Missing Data with Available Complete-Case Missing Value Assumption. https://arxiv.org/abs/2207.02289.

Prepare ACCMV Data

Description

Validates data and identifies observed-value patterns.

Usage

accmv_data(x, y)
single_data_preparation(x, y)
multiple_data_preparation(x, y)

Arguments

Value

An object containing the data and stable missing-pattern encodings.

Examples

x <- matrix(c(1, NA, 2, 3, NA, 4), ncol = 2)
y <- c(1, NA, 2)
accmv_data(x, y)

Bootstrap ACCMV Estimators

Description

Performs a nonparametric row bootstrap, retrying draws lacking estimable pattern comparisons.

Usage

bootstrap_accmv_single(data, method = "mr", n_B = 999, seed = NULL, ...)
bootstrap_accmv_multiple(data, method = "mr", target = "average",
  threshold = NULL, n_B = 999, seed = NULL, delta = NULL)
single_bootstrap(data, method, n_B = 1000, ...)
multiple_bootstrap(data, method, n_B = 1000, ...)
bootstrap_regression(data, method, fm, n_B = 1000, ...)

Arguments

Value

A numeric vector of bootstrap estimates.

ACCMV Estimators

Description

Fits the estimators described in Cheng et al. (2022).

Usage

estimate_accmv_single(x, y, method = "mr", fun = identity,
  binary = FALSE, delta = NULL, n_boot = 0, level = 0.95, seed = NULL)
estimate_accmv_multiple(x, y, method = "mr", target = "average",
  threshold = NULL, delta = NULL, n_boot = 0, level = 0.95, seed = NULL)
single_regression_adjustment(data, fun = identity, binary = FALSE)
single_ipw(data, fun = identity, delta = NULL)
single_ipw_sensitivity(data, delta, fun = identity)
single_multiply_robust(data, fun = identity, binary = FALSE)
multiple_ipw(data, fun = NULL, target = "average", threshold = NULL,
  delta = NULL)
multiple_ipw_sensitivity(data, delta, fun = NULL, target = "average",
  threshold = NULL)
multiple_ra_average(data)
multiple_ra_product(data)
multiple_ra_indicator(data, a)
multiple_mr_average(data)
multiple_mr_product(data)
multiple_mr_indicator(data, a)
accmv_ipw_weights(data)
ipw_regression(data)

Arguments

Value

High-level functions return an accmv_result; low-level functions return an estimate or weights.

Examples

sim <- simulate_accmv_single(500, seed = 1)
estimate_accmv_single(sim$x, sim$y, method = "ra")

sim2 <- simulate_accmv_multiple(500, seed = 2)
estimate_accmv_multiple(sim2$x, sim2$y, method = "mr", target = "product")

Fit an ACCMV-Weighted Marginal Linear Model

Description

Fits a weighted least-squares marginal model using the ACCMV IPW weights from Section 5.

Usage

fit_accmv_regression(x, y, response = 2, predictors = 1,
  n_boot = 0, level = 0.95, seed = NULL)

Arguments

Value

An accmv_regression_result with coefficients, weights, and optional bootstrap inference.

Examples

sim <- simulate_accmv_regression(500, seed = 3)
fit_accmv_regression(sim$x, sim$y)

Print an ACCMV Regression Result

Description

Prints the ACCMV-weighted regression coefficients.

Usage

## S3 method for class 'accmv_regression_result'
print(x, ...)

Arguments

Value

The input object, invisibly.

Print an ACCMV Result

Description

Prints the method, target, point estimate, and bootstrap standard error when available.

Usage

## S3 method for class 'accmv_result'
print(x, ...)

Arguments

Value

The input object, invisibly.

Paper Simulation Designs

Description

Generates data from Sections 7.1 and 7.2 of the ACCMV paper.

Usage

simulate_accmv_single(n = 2000, seed = NULL)
simulate_accmv_multiple(n = 2000, seed = NULL)
simulate_accmv_regression(n = 2000, seed = NULL)

Arguments

Value

A list containing x and y.