| Type: | Package |
| Title: | A Partitioned Quasi-Likelihood for Distributed Statistical Inference |
| Version: | 0.1.0 |
| Author: | Guangbao Guo [aut, cre], Jiarui Li [aut] |
| Maintainer: | Guangbao Guo <ggb11111111@163.com> |
| Description: | In the big data setting, working data sets are often distributed on multiple machines. However, classical statistical methods are often developed to solve the problems of single estimation or inference. We employ a novel parallel quasi-likelihood method in generalized linear models, to make the variances between different sub-estimators relatively similar. Estimates are obtained from projection subsets of data and later combined by suitably-chosen unknown weights. The philosophy of the package is described in Guo G. (2020) <doi:10.1007/s00180-020-00974-4>. |
| License: | MIT + file LICENSE |
| Encoding: | UTF-8 |
| Imports: | parallel,pracma |
| Suggests: | testthat (≥ 3.0.0) |
| Config/testthat/edition: | 3 |
| NeedsCompilation: | no |
| Packaged: | 2024-05-16 11:30:36 UTC; LJR |
| Repository: | CRAN |
| Date/Publication: | 2024-05-21 14:20:02 UTC |
The weight Gauss-Newton estimators of the PQL in Poisson-GLMS
Description
The average weighted estimator and the unknown weighted estimator of the PQL in Poisson-GLMS through damped Gauss-Newton
Usage
pqlBpoisson1(data,G,nk)Arguments
data |
is a design matrix with uniform distribution and the response vector |
G |
is the number of subsets. |
nk |
is the size of subsets |
Value
betaBA, betaBW, MSEA, MSEW
Examples
G <- 20;n=1000;p=5; nk=50
X<- matrix(runif(1000* 5, 0, 0.5), ncol = 5)
beta =matrix(runif(p, 0, 1),nrow=p)
L=X%*%beta
y<- rpois(1000, exp(L))
data=cbind(y,X)
pqlBpoisson1(data,G,nk)The weighted Gauss-Newton estimators of the PQL in Logistic-GLMs
Description
The average weighted estimator and the unknown weighted estimator of the PQL in Logistic-GLMs through damped Gauss-Newton updates.
Usage
pqlBLogist(data,G,nk)
Arguments
data |
is a design matrix with uniform distribution and the response vector. |
G |
is the number of subsets. |
nk |
is the size of subsets. |
Value
betaBW,betaBA,MSEW,MSEA
Examples
G <- 20;n=1000;p=5; nk=50
b=runif(p, 0, 1)
beta =matrix(b,nrow=p)
X=matrix(rnorm(n*p),nrow=n)
L=X%*%beta
prob=1/exp(-(0.48+(L))+1)
y=1/(1+exp(-X))
y=(prob>runif(n))
y= ifelse((prob>runif(n)), 1, 0)
data=cbind(y,X)
pqlBLogist(data,G,nk)
The weighted Gauss-Newton estimators of the PQL in Poisson-GLMS
Description
The average weighted estimator and the unknown weighted estimator of the PQL in Poisson-GLMS through damped Gauss-Newton
Usage
pqlBpoisson2(data,G,nk)Arguments
data |
is a design matrix with uniform distribution and the response vector |
G |
is the number of subsets. |
nk |
is the size of subsets. |
Value
betaBA, betaBW, MSEA, MSEW
Examples
p<- 5;G<- 20;n<- 1000;nk=50
X<- matrix(runif(n * p, 0, 0.5), ncol = p)
beta =matrix(runif(p, 0, 1),nrow=p)
L=X%*%beta
y<- rpois(n, exp(L))
data=cbind(y,X)
pqlBpoisson2(data,G,nk)
pqlLogist
Description
The average weighted estimator and the unknown weighted estimator of the PQL in Poisson-GLMS through damped Gauss-Newton
Usage
pqlLogist(data,G,nk)Arguments
data |
data is a highly correlated data set |
G |
G is the number of nodes |
nk |
n1 is the length of each data subset |
Value
betaW |
estimation value of betaW |
betaA |
estimation value of betaA |
MSEW |
estimation of MSEW |
MSEA |
estimation of MSEA |
Examples
p<- 5;G<- 20;n<- 1000;nk=200
X<- matrix(runif(n*p, 0, 0.5), ncol = p)
beta =matrix(runif(p, 0, 1),nrow=p)
L=X%*%beta
y<- rpois(n, exp(L))
data=cbind(y,X)
pqlLogist(data,G,nk)
The weighted Gauss-Newton estimators of the PQL in Poisson-GLMs
Description
The average weighted estimator and the unknown weighted estimator of the PQL in Poisson-GLMS through damped Gauss-Newton
Usage
pqlPoisson(data,G,nk)Arguments
data |
is a design matrix with uniform distribution and the response vector |
G |
is the number of subsets |
nk |
is the number of outer subsets. |
Value
betaBA, betaBW, MSEA, MSEW
Examples
#library(parallel)
#library(numDeriv)
#library(Rmpi)
#install.packages("pracma");
#library(pracma)
p<- 5;G<- 20;n<- 1000;nk=200
X<- matrix(runif(n*p, 0, 0.5), ncol = p)
beta =matrix(runif(p, 0, 1),nrow=p)
L=X%*%beta
y<- rpois(n, exp(L))
data=cbind(y,X)
pqlPoisson(data,G,nk)