bal.fe.psa: Fisher's Exact Test for Independence of Treatments within...

In PSAgraphics: Propensity Score Analysis Graphics

Fisher's Exact Test for Independence of Treatments within Strata

Description

Simple function that calls fisher.test repeatedly for each strata, testing the independence of treatements for the given covariate within strata.

Usage

bal.fe.psa(categorical, treatment = NULL, strata = NULL, FB = 2000)

Arguments

categorical	Categorical covariate that is being balanced within strata in a PSA. If categorical has three columns, then the second and third are assumed to be the treatment and strata respectively. Missing values are not allowed. May be factor or numeric.
treatment	Binary variable of same length as categorical; generally 0 for 'control,' 1 for 'treatment.'
strata	Integer variable; a vector of same length as categorical indicating the derived strata from estimated propensity scores.
FB	Numeric; number of replications sent to fisher.test.

Details

This function makes repeated calls to fisher.test, Fisher's Exact test, to test whether the distribution of the covariate categorical is independent of treatment within each stratum; a list of p-values for the test for each stratum are returned.

Value

Returns list of the same lenght as the number of strata containing p-values for the indpendence of treatment within each stratum derived from Fisher's Exact test.

Author(s)

James E. Helmreich [e-mail]

Robert M. Pruzek [e-mail]

See Also

bal.cs.psa, bal.ms.psa, bal.ks.psa

Examples

#Everything random
categorical<-sample(4, 1000, replace = TRUE)
treatment<-sample(c(0,1), 1000, replace = TRUE)
strata<-sample(5, 1000, replace = TRUE)
bal.fe.psa(categorical, treatment, strata)

#Perfect balance on 80%, random on last 20%
categorical<-rep(sample(5,1000, replace=TRUE), 2)
treatment<-c(rep(0,1000), rep(1,1000))
strata<-sample(6, 1200, replace=TRUE)
strata<-c(strata[1:1000], strata[1:800], strata[1001:1200])
bal.fe.psa(categorical, treatment, strata)

PSAgraphics documentation built on March 31, 2023, 5:30 p.m.