stepVIF: Stepwise model selection using AIC/QIC for lm, glm/geeglm...

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

Description

Stepwise model selection using AIC/QIC for lm, glm/geeglm objects.

Usage

stepVIF(
  object,
  scope,
  direction = "forward",
  VIFThreshold = 5,
  parallel = FALSE,
  cl.type = NULL,
  n.cores = NULL,
  verbose = TRUE,
  AUC = FALSE,
  ...
)

Arguments

object	A lm, glm, or geeglm object. If the object is defined, the scope argument should be a formula.
scope	Defines the range of models explored in the stepwise search. See details for how to calibrate the scope.
direction	The direction (mode) of the stepwise model. Only "forward"" direction is supported.
VIFThreshold	maximum VIF threshold. Default value is 5.
parallel	If TRUE, run the function in parallel mode.
cl.type	Cluster type ("FORK" or "PSOCK") in parallel mode. Default value is "PSOCK." See parallel package's documentation for more information.
n.cores	Number of cores in parallel mode. Default value is the total number of cores (including logical cores) minus 1.
verbose	If true, prints the steps of the model selection. Default value is FLASE.
AUC	If true, computes the model's AUC along with AIC/QIC (binary reponse only).
...	See lm/glm/gee packages' documentations 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<-lm(log(creatinine) ~ 1,data=donor.kidney.distinct)
summary(fitLM)

frm<-as.formula(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<-stepVIF(object=fitLM,scope=frm)

fitGLM<-with(donor.kidney.distinct, glm(discard ~ 1, family =  binomial(link="logit")))
frm<-as.formula(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)

fitGLM<-stepVIF(object=fitGLM,scope=frm, AUC= FALSE)
summary(fitGLM)

library(geepack)
fit.gee<-geepack::geeglm(discard ~ 1, family =  binomial(link="logit"), data=donor.kidney,
corstr="independence", id=id)
fitGEE<-stepVIF(object=fit.gee,scope=frm, AUC = TRUE, verbose = FALSE)
summary(fitGEE)

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