RenewGLM.outloop: Renewable GLM function processing data batches externally

In luolsph/RenewGLM_pkg: Renewable Estimation and Incremental Inference in Generalized Linear Models with Streaming Datasets

Renewable GLM function processing data batches externally

Description

Take in data batches sequentially and update the regression coefficients and their standard errors

Usage

RenewGLM_out(X, y, type, betahat, infomats, intercept, s, phi)

Arguments

X	covariate matrix for the current data batch
y	response vector for the current data batch
type	the GLM family you want to fit your data to c("gaussian","binomial","poisson")
betahat	the old estimates that need to be updated
infomats	the old cumulative information matrix that need to be updated
intercept	if an intercept is included in the model
s	the cumulative sample size (only needs to be specified in Gaussian model, does not include the samples in the current data batch)
phi	the old estimate of the dispersion parameter in Gaussian model

Value

updated coefficient estimates and the cumulative information matrix

Examples

N=1000
B=10
p=5
n=N/B
beta<-c(0.2,-0.2,0.2,-0.2,0.2)
infomats<-diag(0,p,p);
betahat<-rep(0,p)
for(b in 1:10){
 data<-datagenerator_out(beta,b,n,"binomial","cs",0.5)
 y<-data[,1]
 X<-data[,-1]
 summary<-RenewGLM_out(X,y,"binomial",betahat,infomats,intercept=TRUE,s,phi)
 betahat<-summary[[1]]
 infomats<-summary[[2]]
 rm(data)
}
sd<-sqrt(diag(solve(infomats)))
pvalue<-2*pnorm(-abs(betahat)/sd)
result<-cbind(betahat=betahat,sd=sd,pvalue=pvalue)
colnames(result)<-c("Estimates","Std.Errors","p-values")

luolsph/RenewGLM_pkg documentation built on April 17, 2023, 12:36 a.m.