get_gamma_EB: Calculate the empirical Bayes (EB) estimates

In MetaIntegration: Ensemble Meta-Prediction Framework

Description

Calculate the empirical Bayes (EB) estimates

Usage

get_gamma_EB(gamma_I, gamma_CML, asyV.I)

Arguments

gamma_I	Full model parameter estimates using the internal data only (MLE from direct regression)
gamma_CML	Full model parameter estimates using the internal data and the external reduced model parameters (Chatterjee et al. 2016)
asyV.I	Variance-covariance matrix of gamma_I from function asympVar_LinReg() or asympVar_LogReg()

Value

a list with:

• "gamma_I" Full model parameter estimates using the internal data only (MLE from direct regression)

• "gamma_CML" Full model parameter estimates using the internal data and the external reduced model parameters (Chatterjee et al. 2016)

• "gamma_EB" The empirical Bayes estimate of the full model (i.e. a weighted average of gamma_I and gamma_CML) (Estes et al. 2017)

References

#' Chatterjee, N., Chen, Y.-H., P.Maas and Carroll, R. J. (2016). Constrained maximum likelihood estimation for model calibration using summary-level information from external big data sources. Journal of the American Statistical Association 111, 107-117.

Gu, T., Taylor, J.M.G. and Mukherjee, B. (2020). An ensemble meta-prediction framework to integrate multiple regression models into a current study. Manuscript in preparation.

Examples

# Full model: Y|X1, X2, B
# Reduced model 1: Y|X1 of sample size m1
# Reduced model 2: Y|X2 of sample size m2
# (X1, X2, B) follows normal distribution with mean zero, variance one and 
#  correlation 0.3
# Y|X1, X2, B follows Bernoulli[expit(-1-0.5*X1-0.5*X2+0.5*B)], 
#  where expit(x)=exp(x)/[1+exp(x)]
set.seed(2333)
n = 1000
data.n = data.frame(matrix(ncol = 4, nrow = n))
colnames(data.n) = c('Y', 'X1', 'X2', 'B')
data.n[,c('X1', 'X2', 'B')] = MASS::mvrnorm(n, rep(0,3), diag(0.7,3)+0.3)
data.n$Y = rbinom(n, 1, expit(-1 - 0.5*data.n$X1 - 0.5*data.n$X2 + 0.5*data.n$B))

# Generate the beta estimates from the external reduced model:
# generate a data of size m from the full model first, then fit the reduced regression 
# to obtain the beta estiamtes and the corresponsing estimated variance 
m = m1 = m2 = 30000
data.m = data.frame(matrix(ncol = 4, nrow = m))
names(data.m) = c('Y', 'X1', 'X2', 'B')
data.m[,c('X1', 'X2', 'B')] = MASS::mvrnorm(m, rep(0,3), diag(0.7,3)+0.3)
data.m$Y = rbinom(m, 1, expit(-1 - 0.5*data.m$X1 - 0.5*data.m$X2 + 0.5*data.m$B))

#fit Y|X to obtain the external beta estimates, save the beta estimates and the 
# corresponding estimated variance 
fit.E1 = glm(Y ~ X1, data = data.m, family = binomial(link='logit'))
fit.E2 = glm(Y ~ X2, data = data.m, family = binomial(link='logit'))
beta.E1 = coef(fit.E1)
beta.E2 = coef(fit.E2)
names(beta.E1) = c('int', 'X1')
names(beta.E2) = c('int', 'X2')
V.E1 = vcov(fit.E1)
V.E2 = vcov(fit.E2)

#Save all the external model information into lists for later use
betaHatExt_list = list(Ext1 = beta.E1, Ext2 = beta.E2)
CovExt_list = list(Ext1 = V.E1, Ext2 = V.E2)
rho = list(Ext1 = n/m1, Ext2 = n/m2)

#get full model estimate from direct regression using the internal data only
fit.gamma.I = glm(Y ~ X1 + X2 + B, data = data.n, family = binomial(link='logit'))
gamma.I = coef(fit.gamma.I)

#Get CML estimates using internal data and the beta estimates from the external 
# model 1 and 2, respectively
gamma.CML1 = fxnCC_LogReg(p=2, q=4, YInt=data.n$Y, XInt=data.n$X1, 
                          BInt=cbind(data.n$X2, data.n$B), betaHatExt=beta.E1, 
                          gammaHatInt=gamma.I, n=nrow(data.n), tol=1e-8, 
                          maxIter=400,factor=1)[["gammaHat"]]
gamma.CML2 = fxnCC_LogReg(p=2, q=4, YInt=data.n$Y, XInt=data.n$X2, 
                          BInt=cbind(data.n$X1, data.n$B), betaHatExt=beta.E2, 
                          gammaHatInt=gamma.I, n=nrow(data.n), tol=1e-8, 
                          maxIter=400, factor=1)[["gammaHat"]]
#It's important to reorder gamma.CML2 so that it follows the order (X1, X2, X3, B) 
# as gamma.I and gamma.CML1
gamma.CML2 = c(gamma.CML2[1], gamma.CML2[3], gamma.CML2[2], gamma.CML2[4])

#Get Variance-covariance matricx of c(gamma.I, gamma.CML1, gamma.CML2)
asy.CML = asympVar_LogReg(k=2, p=2,q=4, YInt=data.n$Y, XInt=data.n[,c('X1','X2')], 
                          BInt=data.n$B,  gammaHatInt=gamma.I, 
                          betaHatExt_list=betaHatExt_list, CovExt_list=CovExt_list, 
                          rho=rho, ExUncertainty=TRUE)
asyV.I = asy.CML[["asyV.I"]]                     

#Get the empirical Bayes (EB) estimates
gamma.EB1 = get_gamma_EB(gamma_I=gamma.I, gamma_CML=gamma.CML1, asyV.I=asyV.I)[['gamma.EB']]
gamma.EB2 = get_gamma_EB(gamma_I=gamma.I, gamma_CML=gamma.CML2, asyV.I=asyV.I)[['gamma.EB']]

MetaIntegration documentation built on March 18, 2021, 1:06 a.m.