glmnetVIF: fit a GLM/LM with lasso or elasticnet regularization when VIF...

In mbarah/stepVIF: Stepwise model selection when Varaible Inflation Factor (VIF) is a metric of interest

Description

fit a GLM/LM with lasso or elasticnet regularization when VIF is also a metric of interest

Usage

glmnetVIF(object, VIFThreshold = 5, cv.lambda = "min", verbose = FALSE, ...)

Arguments

object	An lm or glm object.
VIFThreshold	maximum VIF threshold. Default value is 5.
cv.lambda	c("min","1se") the tuning factor for lambda in glmnet cross validation. Default value is "min". See glmnet documentation for more information.
verbose	If true, prints the steps of the model selection. Default value is FALSE.
...	See glmnet package's documentation for additional parameters.

Author(s)

Masoud Barah, Sanjay Mehrotra

Examples

data("donor.kidney")
#select distinct donors
donor.kidney.distinct<-donor.kidney[!duplicated(donor.kidney[c("id")]),]

fitLM<-with(donor.kidney, lm(log(creatinine) ~ log(KDRI) + glomerulosclerosis +race+
                  anti_HCV + on_pump + DCD + laterality +
                  init_pump_resistance + terminal_pump_resistance +
                  init_pump_flow+ diabetes + smoking +
                  blood_type + HBsAg + MI + clinical_infection +
                  anti_HBs + Tattoos + cancer +
                  CMV + anti_HBc + HTLV))
fitLM<-glmnetVIF(object=fitLM)

fitGLM<-glm(discard ~ log(KDRI) + log(creatinine)+ glomerulosclerosis +race+
                  anti_HCV + on_pump + DCD + laterality +
                  init_pump_resistance + terminal_pump_resistance +
                  init_pump_flow+ diabetes + smoking +
                  blood_type + HBsAg + MI + clinical_infection +
                  anti_HBs + Tattoos + cancer +
                  CMV + anti_HBc + HTLV,family =  binomial(link="logit"),data=donor.kidney)
summary(fitGLM)

fitGLM<-glmnetVIF(object=fitGLM,VIFThreshold = 5, verbose = TRUE)
summary.glm(fitGLM)

mbarah/stepVIF documentation built on Feb. 19, 2021, 8:06 p.m.