PWLS {vsmi}

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")