rbounds (version 0.2)

mcontrol: Sensitivity Analysis For Multiple Matched Controls

Description

Function to calculate Rosenbaum bounds for continuous or ordinal outcomes based on Wilcoxon sign rank test p-value when there are multiple matched control units.

Usage

#Default Method 
mcontrol(y, grp.id, treat.id, group.size=3, Gamma=4, GammaInc=1)

Arguments

y

Vector of grouped matched outcomes.

treat.id

A vector indicating the treated unit in each matched group.

grp.id

A vector indicating matched groups.

group.size

The size of the matched groups. Three for one treated unit and two control units.

Gamma

Upper-bound on gamma parameter.

GammaInc

To set increments for gamma parameter.

Details

The matched data needs to be in a very particular form for this function to work. The data must be sorted by matched groups with indicators for each matched group and for treated and control units. The simplest way to prepare the data is to use the Match() function and use the data.prep() function to format the data.

Currently this function only takes matched data with 2 or 3 controls units matched to each treated unit.

This function does cannot handle data where the number of control units is not the same for every treated unit.

References

Rosenbaum, Paul R. (2002) Observational Studies. Springer-Verlag.

See Also

See also data.prep, binarysens, psens, Match, hlsens

Examples

#
#Load Matching Software and Data
#
library(Matching)
data(lalonde)

#
# Estimate Propensity Score
#
DWglm  <- glm(treat~age + I(age^2) + educ + I(educ^2) + black +
             hisp + married + nodegr + re74  + I(re74^2) + re75 + I(re75^2) +
             u74 + u75, family=binomial, data=lalonde)

#
#save data objects
#
Y  <- lalonde$re78   #the outcome of interest
Tr <- lalonde$treat #the treatment of interest

#
# Match
#             
mDW  <- Match(Y=Y, Tr=Tr, X=DWglm$fitted, M=2)

#
# One should check balance, but let's skip that step for now.
#

#Create Data Object
tmp <- data.prep(mDW, group.size=3)

#
# Sensitivity Test
#
mcontrol(tmp$Y, tmp$id, tmp$treat, group.size=3)