StepReg Vignette

In StepReg: Stepwise Regression Analysis

Introduction

Here we applied StepReg to the well-known mtcars data and lung data for clarifying how to perform linear, logistic and Cox stepwise regression.

#install.package("StepReg")
library(StepReg)

linear stepwise regression with data mtcar

linear stepwise regression using 'forward' method for variable selection and 'AIC' as criteria for stop rules

formula <- mpg ~ .
sForwAIC <- stepwise(formula=formula,
         data=mtcars,
         selection="forward",
         select="AIC")
sForwAIC

From the above result, we can see that stepwise() output a list with 5 tables.

• 'Summary of Parameters' tells you what parameters is used for this function, where Intercept equals to 1 showing that this stepwise regression has a intercept, otherwise 0 has not a intercept.

• 'Variables Type' is the summary of the type of all variables.

• 'Process of Selection' let you know how variables are selected, we used AIC as the criteria so the last column is value of AIC.

• 'Coefficients of the Selected Variables' is the coefficients of all selected variable.

linear stepwise regression using 'bidirection' method for variable selection and 'SL' as criteria for stop rules, and we set significant level for entry(sle) is 0.15, and significant level of stay(sls) is 0.15 too.

formula <- mpg ~ .
sBidiSL <- stepwise(formula=formula,
         data=mtcars,
         selection="bidirection",
         select="SL",
         sle=0.15,
         sls=0.15)
sBidiSL

The output of this time is similar to the last time except that the last column name of 'Process of Selection' is SL.

linear stepwise regression using 'backward' method for variable selection and 'SBC' as criteria for stop rules and without intercept in stepwise regression.

#formula <- mpg ~ . -1
formula <- mpg ~ . + 0
sBackSBC <- stepwise(formula=formula,
                     data=mtcars,
                     selection="backward",
                     select="SBC")
sBackSBC

Note that 0 instead of 1 in table 'Summary of Parameters' and 'Selected Varaibles'.

linear stepwise regression using 'score' method for variable selection, 'AICc' as criteria and we perform multivariable multiple stepwise regression with mpg and dra as dependent variables, besides, we select cyl disp hp wt vs am as independent variables, where wt is included always.

formula <- cbind(mpg,drat) ~ cyl+disp+hp+wt+vs+am
stepwise(formula=formula,
        data=mtcars,
        include='wt',
        selection="score",
        select="AICc")

logistic stepwise regression with data mtcars

logistic stepwise regression using 'forward' method for variable selection and 'AIC' as criteria for stop rules

formula <- am ~ .
stepwiseLogit(formula=formula,
              data=mtcars,
              selection="forward",
              select="AIC")

logistic stepwise regression using 'score' method for variable selection and 'SL' as criteria and only output the first 3 best model.

formula <- am ~ .
stepwiseLogit(formula=formula,
              data=mtcars,
              selection="score",
              select="SL",
              best=3)

Cox stepwise regression with data lung

cox stepwise regression using 'forward' method for variable selection and 'IC(1)' as criteria for stop rules

lung <- survival::lung
my.data <- na.omit(lung)
my.data$status1 <- ifelse(my.data$status==2,1,0)
data <- my.data
formula = Surv(time, status1) ~ . - status 

stepwiseCox(formula=formula,
  data=my.data,
  selection="forward",
  select="IC(1)")

cox stepwise regression using 'score' method for variable selection and 'SL' as criteria and only output the first 3 best model.

formula = Surv(time, status1) ~ . - status 
stepwiseCox(formula=formula,
  data=my.data,
  selection="score",
  select="SL",
  best=3)

Session Info

sessionInfo()

StepReg documentation built on Dec. 28, 2022, 1:07 a.m.