In Efaq/lab4elife599carde734: Linear regression package for assignment 4 in Advanced R

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

Predictive modeling with Ridgereg

Let's do some analysis regarding predictive modeling. The dataset BostonHousing will be used in this demonstration.

library(lab4elife599carde734)
library(caret)
library(mlbench)
library(leaps)

First, we need to divide the BostonHousing data into a test and training dataset.

data("BostonHousing")

set.seed(123)
trainIndex <- createDataPartition(BostonHousing$medv, p = .8, 
                                  list = FALSE, 
                                  times = 1)

BostonTrain <- BostonHousing[ trainIndex,]
BostonTest  <- BostonHousing[-trainIndex,]

head(BostonTrain)
head(BostonTest)

Fit a linear regression model

lm_model <- train(medv ~ .,
               data = BostonTrain,
               method = "lm")

Fit a linear regression model with forward selection of covariates

forward_selection  <- train(medv ~ ., data  =  BostonTrain ,
                             method  =  "leapForward" )

Now we want to evaluate the perfomance of the models that we have just created. To do that, we need to calculate the RMSE (the lower the better) and R-squared (the bigger the better). Both RMSE and R-squared quantify how well a regression model fits a dataset. The RMSE tells us how well a regression model can predict the value of the response variable in absolute terms while R-squared tells us how well a model can predict the value of the response variable in percentage terms

trainX  <-  BostonTrain[,-(which(names(BostonTrain)=="medv"))]
trainY  <-  BostonTrain$medv


testX  <-  BostonTest[,-(which(names(BostonTest)=="medv"))]
testY  <-  BostonTest$medv

predictions_lm <- predict(lm_model, BostonTest)
predictions_forward <- predict(forward_selection, BostonTest)


#RMSE - The lower the better

lm_model_rmse<- RMSE(predictions_lm,testY)
lm_forward_rmse<- RMSE(predictions_forward,testY)

#R2 (R-squared) - The higher the better

lm_model_r2<- R2(predictions_lm,testY)
lm_forward_r2<- R2(predictions_forward,testY)

#RMSE for Linear Regression
print(lm_model_rmse)

#RMSE for Linear Regression forward selection
print(lm_forward_rmse)

#R2 for Linear Regression
print(lm_model_r2)

#R2 for Linear Regression forward selection
print(lm_forward_r2)

#For both indicator, linear regression model seems to be the model that predicts better the data

Efaq/lab4elife599carde734 documentation built on Dec. 17, 2021, 6:23 p.m.