ds_function: Estimating design sensitivity.

In Ngendahimana/sensitivityR5: This package supports implementation of sensitivity analyisis using objects from Match, MatchIt and bmatch

Description

This function estimates design sensitivity parameter associated with a list of matching algorithms.

Usage

ds_function(x, sf = 1/3, Card_Dat = NULL)

Arguments

x	A list of matched sampling algorithms.
sf	A fraction used to split the entire matched sample into planning sample (to estimate design sensitivity) and analysis sample (for outcome analysis).
Card_Dat	Dataframe passed to the function implementing cardinality matching. Passed to ds_function only if cardinality matching is one of the matching algorithms to be evaluated

Examples

# Implementing different types of matching using matchit package
m.out_Nrst <- matchit(exposure~V2+V3+V5+V6+V8+V9,data=dat, method = "nearest", distance = "logit",replace = TRUE ,print = FALSE,ratio = 1)
m.out_NrstCAL <- matchit(exposure~V2+V3+V5+V6+V8+V9,data=dat, method = "nearest", distance = "logit",replace = TRUE,print = FALSE,ratio = 1,caliper=0.2)
m.out_optimal <- matchit(exposure~V2+V3+V5+V6+V8+V9,data=dat, method = "optimal", distance = "logit",replace =TRUE,print = FALSE,ratio = 1)
m.out_genetic <- matchit(exposure~V2+V3+V5+V6+V8+V9,data=dat, method = "genetic", distance = "logit",replace = TRUE,print = FALSE,ratio = 1,unif.seed=1945, int.seed=1906)


# Implementing cardinality matching
# Solver options
t_max = 60*5
solver = "glpk"
approximate = 1
solver = list(name = solver, t_max = t_max, approximate = approximate,round_cplex = 0, trace = 0)
Card_Dat = dat[order(dat$exposure==1, decreasing = TRUE), ]
Card_Dat$`.distance` <- (glm(exposure~V2+V3+V5+V6+V8+V9, data = Card_Dat ,family = binomial(link="logit")))$fitted.values

attach(Card_Dat)
t_ind = exposure
# Distance matrix
dist_mat = NULL
subset_weight = 1

## Setting balance criteria
mom_covs = cbind(.distance,V2,V3,V5,V6,V8,V9)
mom_tols = round(absstddif(mom_covs, t_ind, .1), 2)
mom = list(covs = mom_covs, tols = mom_tols)

## Creating cardinality matching object
m.out_dsnMatch = bmatch(t_ind = t_ind, dist_mat = dist_mat, subset_weight = subset_weight,mom = mom,solver = solver)

# Assessing balance
t_id1 = m.out_dsnMatch$t_id
c_id1 = m.out_dsnMatch$c_id
dsn_bal = meantab(mom_covs, t_ind, t_id1, c_id1)[,6]
Card_Dat_A = rbind(Card_Dat[t_id1,],Card_Dat[c_id1,])
detach(Card_Dat)
arg.list1 = list("a"=m.out_Nrst , "b"=m.out_NrstCAL)
arg.list2= list("a"=m.out_Nrst , "b"=m.out_NrstCAL, "c"=m.out_dsnMatch, "d"=m.out_optimal)
k=ds_function(arg.list2,Card_Dat = Card_Dat)
k$designSensitivity

Ngendahimana/sensitivityR5 documentation built on June 24, 2020, 4:09 a.m.