This package provides functions for Inference for the treatment effect with possibly invalid instrumental variables, including the Two-Stage Hard Thresholding(TSHT) method, the Endogeneity Testing method, and Searching-Sampling method.
The package can be installed from Github using the following code:
devtools::install_github("https://github.com/zijguo/RobustIV")Before using the package, we can use the following code:
library(RobustIV)We use pseudodata provided by Youjin Lee, which is generated mimicing the structure of Framingham Heart Study data. We assume Y is a linear model of D,Z, and X, and D is a linear model of Z and X.
> data("lineardata")
> Y <- lineardata[,"Y"]
> D <- lineardata[,"D"]
> Z <- as.matrix(lineardata[,c("Z.1","Z.2","Z.3","Z.4","Z.5","Z.6","Z.7","Z.8")])
> X <- as.matrix(lineardata[,c("age","sex")])
> TSHT.model <- TSHT(Y=Y,D=D,Z=Z,X=X)
> summary(TSHT.model)
Relevant IVs: Z.3 Z.4 Z.5
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
betaHat Std.Error CI(2.5%) CI(97.5%) Valid IVs
0.05166598 0.02015546 0.012162 0.09116995 Z.3 Z.4 Z.5
Guo, Z., Kang, H., Tony Cai, T. and Small, D.S. (2018), Confidence intervals for causal effects with invalid instruments by using two-stage hard thresholding with voting, J. R. Stat. Soc. B, 80: 793-815.
> Y <- lineardata[,"Y"]
> D <- lineardata[,"D"]
> Z <- as.matrix(lineardata[,c("Z.1","Z.2","Z.3","Z.4","Z.5","Z.6","Z.7","Z.8")])
> X <- as.matrix(lineardata[,c("age","sex")])
> Searching.model <- SearchingSampling(Y,D,Z,X, Sampling = FALSE)
> summary(Searching.model)
Initial set of Valid Instruments: Z.3 Z.4 Z.5
Plurality rule holds.
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Confidence Interval for Beta: [-0.0356797,0.1422332]
> Sampling.model <- SearchingSampling(Y,D,Z,X)
> summary(Sampling.model)
Initial set of Valid Instruments: Z.3 Z.4 Z.5
Plurality rule holds.
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Confidence Interval for Beta: [-0.02297164,0.1295251]Guo, Z. (2021), Causal Inference with Invalid Instruments: Post-selection Problems and A Solution Using Searching and Sampling, Preprint arXiv:2104.06911.
In this section, we consider the following linear models.
Y = D\beta + Z\alpha + X\phi +u
D = Z\gamma + X\psi + v
> set.seed(1)
> n = 500; L = 600; s = 3; k = 10; px = 10;
> alpha = c(rep(3,s),rep(0,L-s)); beta = 1; gamma = c(rep(1,k),rep(0,L-k))
> phi<-(1/px)*seq(1,px)+0.5; psi<-(1/px)*seq(1,px)+1
> epsilonSigma = matrix(c(1,0.8,0.8,1),2,2)
> Z = matrix(rnorm(n*L),n,L)
> X = matrix(rnorm(n*px),n,px)
> epsilon = MASS::mvrnorm(n,rep(0,2),epsilonSigma)
> D = 0.5 + Z %*% gamma + X %*% psi + epsilon[,1]
> Y = -0.5 + Z %*% alpha + D * beta + X %*% phi + epsilon[,2]
> TSHT.model <- TSHT(Y,D,Z,X,method = "Fast.DeLasso")
> summary(TSHT.model)
Relevant IVs: 1 2 3 4 5 6 7 8 9 10
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
betaHat Std.Error CI(2.5%) CI(97.5%) Valid IVs
0.9880546 0.0157771 0.9571321 1.018977 4 5 6 7 8 9 10
It uses same reduced form estimator as TSHT in each setting.
> set.seed(1)
> n = 500; L = 600; s = 3; k = 10; px = 10;
> alpha = c(rep(3,s),rep(0,L-s)); beta = 1; gamma = c(rep(1,k),rep(0,L-k))
> phi<-(1/px)*seq(1,px)+0.5; psi<-(1/px)*seq(1,px)+1
> epsilonSigma = matrix(c(1,0.8,0.8,1),2,2)
> Z = matrix(rnorm(n*L),n,L)
> X = matrix(rnorm(n*px),n,px)
> epsilon = MASS::mvrnorm(n,rep(0,2),epsilonSigma)
> D = 0.5 + Z %*% gamma + X %*% psi + epsilon[,1]
> Y = -0.5 + Z %*% alpha + D * beta + X %*% phi + epsilon[,2]
> endo.test.model <- endo.test(Y,D,Z,X, invalid = TRUE)
> summary(endo.test.model)
Valid Instruments: 4 5 6 7 8 9 10
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Estimated covariance: 0.7690532
Test statistics Q = 13.47976
P-value = 0
'H0 : Sigma12 = 0' is rejected at the significance level 0.05 .Guo, Z., Kang, H., Tony Cai, T. and Small, D.S. (2018), Testing endogeneity with high dimensional covariates, Journal of Econometrics, Elsevier, vol. 207(1), pages 175-187.