Sms 23.11

library(tidyverse)
library(modelr)
library(broom)

# Load the data
data("swiss")
# Inspect the data
sample_n(swiss, 3)

summary(swiss)
#Using Linear Model
Model0 <- lm(Fertility ~1, data = swiss)
Model1 <- lm(Fertility ~., data = swiss)
Model2 <- lm(Fertility ~. -Examination, data = swiss)
summary(Model0)
summary(Model1)
summary(Model2)


cbind(AIC(Model0),AIC(Model1),AIC(Model2))
cbind(BIC(Model0),BIC(Model0),BIC(Model2))

library(modelr)
data.frame(
  R2 = rsquare(Model1, data = swiss),
  RMSE = rmse(Model1, data = swiss),
  MAE = mae(Model1, data = swiss)
)


library(caret)
predictions <- Model1 %>% predict(swiss)
data.frame(
  R2 = R2(predictions, swiss$Fertility),
  RMSE = RMSE(predictions, swiss$Fertility),
  MAE = MAE(predictions, swiss$Fertility)
)

library(broom)
rbind(glance(Model0),glance(Model1),glance(Model2))

install.packages("Rtools")

library(d3heatmap)
library(mplot)
library(pairsD3)
library(Hmisc)
library(mfp)
library(dplyr)
loc <- "https://www4.stat.ncsu.edu/~boos/var.select/diabetes.tab.txt"
dat <- as.matrix(read.table(loc,header=T,sep="\t"))
dim(dat)
data("diabetes", package = "mplot")
# help('diabetes', package='mplot')
str(diabetes)  # structure of the diabetes
summary(diabetes) #Descriptive Statistics
pairs(diabetes)  # traditional pairs plot
round(cor(diabetes),4)
boxplot(diabetes)  # always a good idea to check for gross outliers
pairsD3::shinypairs(diabetes)  # interactive pairs plot of the data set
d3heatmap::d3heatmap(cor(diabetes))
Hmisc::describe(diabetes, digits = 1)  # summary of the diabetes data
#Example Linear Model
Model0 = lm(y ~ 1, data = diabetes)  # Null model
summary(Model0)
Model1 = lm(y ~ ., data = diabetes)  # Full model
summary(Model1)
cbind(AIC(Model0),AIC(Model1))
cbind(BIC(Model0),BIC(Model1))
#backward selecton using BIC and AIC
step.back.bic = step(Model1, direction = "backward", trace = FALSE, k = log(442))
summary(step.back.bic)
step.back.aic = step(Model1, direction = "backward", trace = FALSE, k = 2)
summary(step.back.aic)
#Example Generalised Linear Model (GLM)
Model00 = glm(y ~ 1, data = diabetes)  # Null model
Model11 = glm(y ~ ., data = diabetes)  # Full model BY default Gaussian
summary(Model11)
summary(Model00)
Model22 <- glm(y ~ ., data = diabetes, family = "gaussian")
summary(Model22)
step.back.aic = step(Model22, direction = "backward", trace = FALSE, k = 2)
summary(step.back.aic)
anova(Model22, test = "Chisq")