Type: Package
Title: Stochastic Augmentation of Matched Data Using Restriction Methods
Version: 1.1
Date: 2022-08-30
Author: Mansour T.A. Sharabiani, Alireza S. Mahani
Maintainer: Alireza S. Mahani <alireza.s.mahani@gmail.com>
Description: Augmenting a matched data set by generating multiple stochastic, matched samples from the data using a multi-dimensional histogram constructed from dropping the input matched data into a multi-dimensional grid built on the full data set. The resulting stochastic, matched sets will likely provide a collectively higher coverage of the full data set compared to the single matched set. Each stochastic match is without duplication, thus allowing downstream validation techniques such as cross-validation to be applied to each set without concern for overfitting.
License: GPL-2 | GPL-3 [expanded from: GPL (≥ 2)]
Imports: Matching
NeedsCompilation: no
Packaged: 2022-08-30 14:00:19 UTC; ubuntu
Repository: CRAN
Date/Publication: 2022-08-31 13:00:14 UTC

Stochastic Augmentation of Matched Datasets Using Restriction Methods

Description

This function generates multiple subsets of the data in which the distribution of covariates is balanced across treatment groups. It works by binning the output of a base matching algorithm into a multidimensional histogram, and drawing - without replacement - from the full data set according to the histogram. This leads to higher data coverage across multiple matched subsets without duplication of cases within each subset.

Usage

samur(
  formula, data
  , matched.subset = 1:nrow(data)
  , nsmp = 100
  , use.quantile = TRUE, breaks = 10
  , replace = length(unique(matched.subset)) < length(matched.subset)
  )
## S3 method for class 'samur'
print(x, ...)

Arguments

formula

Formula expression used to describe the treatment variable (lhs) and covariates used during matching (rhs).

data

Data frame containing the treatment variables and matched covariates as specified in the formula.

matched.subset

An integer vector representing the indexes of a subset of data that is the output of a base matching algorithm. It cannot contain duplicate values.

nsmp

Number of stochastically matched subsets to generate.

use.quantile

Should numeric covariates be binned using quantiles (TRUE) or not.

breaks

number of breaks to use in binning numeric covariates.

replace

Boolean flag indicating whether or not to perform sampling with replacement.

x

An object of class samur, typically the output of function samur.

...

Arguments passed to/from other methods.

Value

An object of class samur, a matrix of size length(matched.subset) by nsmp, where each column is a matched subset wihtout case duplication. It also has the following attributes:

call

Copy of function call.

formula

Formula passed to the function.

mdg

Multi-dimensional grid used for binning the matched data subsets.

mdh

Multi-dimensional histogram resulting frm binning data[matched.subset, ] according to the grid specified in mdg.

data

Copy of data frame passed to the function.

Author(s)

Mansour T.A. Sharabiani, Alireza S. Mahani

See Also

summary.samur

Examples

## Not run: 
library(SAMUR)
library(Matching)
data(lalonde)
myformula <- treat ~ age + educ
myglm <- glm(myformula, lalonde, family="binomial")
X <- myglm$fitted.values
# using M=1 and replace=F to ensure no duplication
bimatch <- Match(Tr = lalonde$treat, X = myglm$fitted.values
  , M = 1, replace = F, caliper = 0.25)
idx <- c(bimatch$index.control, bimatch$index.treated)
my.samur <- samur(formula = myformula, data = lalonde
  , matched.subset = idx, nsmp = 100
  , breaks = 10, use.quantile = TRUE)
summary(my.samur, nboots = 500)

## End(Not run)

Summarizing Output of SAMUR Augmentation Function

Description

summary method for class "samur".

Usage

## S3 method for class 'samur'
summary(object, ...)
## S3 method for class 'summary.samur'
print(x, ...)

Arguments

object

An object of class "samur", usually the result of a call to samur.

x

An object of class "summary.samur", usually the result of a call to summary.samur.

...

Further arguments to be passed to/from other methods. Current implementation of summary.samur passes arguments to MatchBalance function from Matching package.

Value

A list with the following elements:

min.pval.new

A vector of length equal to number of samples (nsmp) generated by samur, each representing the minimum p-value from all univariate tests performed by the underlying function MatchBalance. It also has an attributed named min.pval.orig, containing a similar number for the original matched subset, i.e. data[matched.subset, ].

min.pval.orig

Same number as above, but for original matched subset.

coverage.new

Percent of cases from full data set covered among all stochastic, matched samples.

coverage.orig

Same as above, calculated for the original matched subset.

Note

All t-tests used for p-value calculations are "not" paired, since the philosophy of stochastic augmentation relaxes the notion of one-to-one matching.

Author(s)

Alireza S. Mahani, Mansour T.A. Sharabiani

See Also

samur, MatchBalance