| Title: | Variable Selection for Multiple Imputed Data |
| Version: | 0.1.0 |
| Description: | Penalized weighted least-squares estimate for variable selection on correlated multiply imputed data and penalized estimating equations for generalized linear models with multiple imputation. Reference: Li, Y., Yang, H., Yu, H., Huang, H., Shen, Y*. (2023) "Penalized estimating equations for generalized linear models with multiple imputation", <doi:10.1214/22-AOAS1721>. Li, Y., Yang, H., Yu, H., Huang, H., Shen, Y*. (2023) "Penalized weighted least-squares estimate for variable selection on correlated multiply imputed data", <doi:10.1093/jrsssc/qlad028>. |
| License: | MIT + file LICENSE |
| Encoding: | UTF-8 |
| RoxygenNote: | 7.3.1 |
| Imports: | MASS (≥ 7.3-60), Matrix (≥ 1.6-1.1), mice (≥ 3.16.0), qif (≥ 1.5) |
| NeedsCompilation: | no |
| Packaged: | 2024-05-24 01:42:43 UTC; 86188 |
| Author: | Mingyue Zhang [aut], Yang Li [aut], Haoyu Yang [aut, cre] |
| Maintainer: | Haoyu Yang <haoyuyang@alu.ruc.edu.cn> |
| Repository: | CRAN |
| Date/Publication: | 2024-05-25 17:20:02 UTC |
vsmi: Variable selection for multiple imputed data
Description
This is a package to implementation penalized weighted least-squares estimate for variable selection on correlated multiply imputed data and penalized estimating equations for generalized linear models with multiple imputation.
Functions
PEE:Penalized estimating equations for generalized linear models with multiple imputation
PWLS : Penalized weighted least-squares estimate for variable selection on correlated multiply imputed data
generate_pwls_missing_data : Generate example missing data for PWLS
generate_pee_missing_data : Generate example missing data for PEE
Author(s)
Maintainer: Haoyu Yang haoyuyang@alu.ruc.edu.cn
Authors:
Mingyue Zhang
Yang Li
Penalized estimating equations for generalized linear models with multiple imputation
Description
This is a function to impute missing data, estimate coefficients of generalized linear models and select variables for multiple imputed data sets, considering the correlation of multiple imputed observations.
Usage
PEE(
missdata,
mice_time = 5,
penalty,
lamda.vec = seq(1, 4, length.out = 12),
Gamma = c(0.5, 1, 1.5)
)
Arguments
missdata |
A Matrix,missing data with variables X in the first p columns and response Y at the last column. |
mice_time |
an integer, number of imputation. |
penalty |
The method for variable selection,choose from "lasso" or "alasso". |
lamda.vec |
Optimal tuning parameter for penalty,default seq(1,4,length.out=12). |
Gamma |
Parameter for adjustment of the Adaptive Weights vector in adaptive LASSO,default c(0.5,1,1.5). |
Value
A Vsmi_est object, contians estcoef and index_sig , estcoef for estimate coefficients and index_sig for selected variable index.
Examples
library(MASS)
library(mice)
library(qif)
data_with_missing <- generate_pee_missing_data(outcome="binary")
est.alasso <-PEE(data_with_missing,penalty="alasso")
est.lasso <-PEE(data_with_missing,penalty="lasso")
count_data_with_missing <- generate_pee_missing_data(outcome="count")
count_est.alasso <-PEE(data_with_missing,penalty="alasso")
count_est.lasso <-PEE(data_with_missing,penalty="lasso")
Penalized weighted least-squares estimate for variable selection on correlated multiply imputed data
Description
This is a functions to estimate coefficients of wighted leat-squares model and select variables for multiple imputed data sets ,considering the correlation of multiple imputed observations.
Usage
PWLS(
missdata,
mice_time = 5,
penalty = "alasso",
lamda.vec = seq(6, 24, length.out = 40),
Gamma = c(0.5, 1, 2)
)
Arguments
missdata |
A Matrix,missing data with variables X in the first p columns and response Y at the last column. |
mice_time |
An intedevger, number of imputation. |
penalty |
The method for variable selection,choose from "lasso" or "alasso". |
lamda.vec |
Optimal tuning parameter for penalty,default seq(1,4,length.out=12). |
Gamma |
Parameter for adjustment of the Adaptive Weights vector in adaptive LASSO,default c(0.5,1,1.5). |
Value
A Vsmi_est object, contians estcoef and index_sig , estcoef for estimate coefficients and index_sig for selected variable index.
Examples
library(MASS)
library(mice)
library(qif)
entire<-generate_pwls_missing_data()
est_lasso<-PWLS(entire,penalty="lasso")
est_alasso <- PWLS(entire,penalty = "alasso")
Generate example data for PEE
Description
This is a functoin to generate example missing data for PEE
Usage
generate_pee_missing_data(
outcome = "binary",
p = 20,
n = 200,
pt1 = 0.5,
tbeta = c(3/4, (-3)/4, 3/4, (-3)/4, 3/4, (-3)/4, (-3)/4, 3/4),
miss_sig = c(1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0)
)
Arguments
outcome |
The type of response variable Y, choose "binary" for binary response or "count" for poisson response,defualt "binary" |
p |
The dimension of the independent variable X,default 20. |
n |
The Number of rows of generated data,default 200. |
pt1 |
Missing rate of independent variable X,default 0.5. |
tbeta |
True value of the coefficient,default c(3/4,(-3)/4,3/4,(-3)/4,3/4,(-3)/4,(-3)/4,3/4). |
miss_sig |
A 0-1 vector of length p, where 1 means that variable at the index is with missing,while 0 means that it without missing,defualt c(1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0) |
Value
A Matrix,missing data with variables X in the first p columns and response Y at the last column.
Generate example data for PWLS
Description
This is a functoin to generate example missing data for PWLS
Usage
generate_pwls_missing_data(
p = 20,
n = 200,
pt1 = 0.5,
pt2 = 0.5,
tbeta = c(1, -1, 1, -1, 1, -1, -1, 1),
miss_sig = c(0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0)
)
Arguments
p |
The dimension of the independent variable X,default 20. |
n |
The Number of rows of generated data,default 200. |
pt1 |
Missing rate of independent variable X,default 0.5. |
pt2 |
Missing rate of response Y, default 0.5. |
tbeta |
True value of the coefficient,default c(1,-1,1,-1,1,-1,-1,1). |
miss_sig |
A 0-1 vector of length p, where 1 means that variable at the index is with missing,while 0 means that it without missing,defualt c(0,1,0,0,1,0,0,0,0,0,1,0,0,0,1,0,0,0,0,0) |
Value
A Matrix,missing data with variables X in the first p columns and response Y at the last column.