BARTpred: A helper function. Allows you to fit observed joint...

In demetrios1/Causallysensitive: What the Package Does (One Line, Title Case)

A helper function. Allows you to fit observed joint probabilities with BART, or with a monotonicity constraint

Description

A helper function. Allows you to fit observed joint probabilities with BART, or with a monotonicity constraint

Usage

BARTpred(
  df,
  treat = "G",
  Outcome = "B",
  vars,
  mono = T,
  nd_post = 2000,
  n_skip = 2000
)

Arguments

df	dataframe including the covariates plus the treatment and outcome columns
treat	String of name of treatment column
Outcome	string, name of outcome column
vars	list of names of columns of covariates
mono	boolean, whether or not to impose monotonicity constraint, use if monbart is installed
nd_post	number of posterior draws to keep
n_skip	number of burn in samples

Examples

#' N <- 500 # Number of random samples
a=1

x1=runif(N, -a,a)
x2=runif(N, -a,a)
x3=runif(N,-a,a)
x4=runif(N,-a,a)
x5=runif(N, -a,a)
beta1= -0.2
alpha1= 0.7
beta0= -0
alpha0= -0.5
mu1 <- beta0+beta1*(x1+x2+x3+x4+x5)
mu2 <- alpha0+alpha1*(x1+x2+x3+x4+x5)
mu<-matrix(c(mu1, mu2), nrow=N, ncol=2)
rho=.5
gamma=1
B1.true=pnorm(mu2+gamma)
B0.true=pnorm(mu2)
sigma <- matrix(c(1, rho,rho,1),
               2) # Covariance matrix
sim_data=t(sapply(1:N, function(i)MASS::mvrnorm(1, mu = mu[i,], Sigma = sigma )))
#generate the binary treatments
G=sapply(1:N, function(i)ifelse(sim_data[i,1]>=0, 1,0))
#generate the binary outcomes
B=sapply(1:N, function(i)ifelse(sim_data[i,2]>=-1*gamma*G[i], 1,0))
print(table(G,B))
covariates=data.frame(x1,x2,x3,x4,x5,B, G)
vars=c('x1','x2','x3','x4','x5')
CV_pred=BARTpred_CV(covariates, treat='G', Outcome='B', vars=vars,mono=T,
 nd_post=200, n_skip=500, nfold=5)

demetrios1/Causallysensitive documentation built on Nov. 18, 2022, 5:27 p.m.