Basics of R- Session 16- Discriminant Analysis

discriminant analysis for two group

rm(list=ls())

#Discriminat analysis

#import data set

da1<-read.csv("C:/Users/Administrator.vaio1/Desktop/discriminant two group hair.csv")
fix(da1)

library(MASS)
library(psych)
library(Hmisc)
library(stats)
library(HSAUR)
library(dawai)
#library(rattle)
library(car)
library(MVN)
library(perturb)
library(biotools)
library(FactoMineR)
library(DiscriMiner)


#create a subset of the data for simplicity say x2 and x6, x7,x8,x9,x10

dasub1<-da1[,c(3,7:19)]
fix(dasub1)

#Henze-Zirkler's MVN test

library(MVN)

multinormality<-mvn(dasub1, subset = "x4", mvnTest = "hz")  

# mvnTest "mardia" for Mardia’s test, 

# "hz" for HenzeZirkler’s test, 

# "royston" for Royston’s test, 

# "dh" for Doornik-Hansen’s test

# and energy for E-statistic

multinormality$multivariateNormality

# outliers can also be identified

multinormality<-mvn(dasub1, subset = "x4", mvnTest = "hz", showOutliers = TRUE)  

multinormality$multivariateOutliers

#multicollinearity

#Homogeneity of variance-covariance matrix using Anderson's test, Box's M test

#boxplot
boxplot(dasub1$x6~dasub1$x2)

#box M test
library(biotools)

# box m (IV, D-group)
boxM(dasub1[,-1],dasub1$x2)

# wilks lambda
library(rrcov)
Wilks.test(dasub1[,-1],dasub1$x2)

or

Wilks.test(dasub1[,-1], dasub1$x4)

Wilks.test(x4~., data = dasub1)

# eign value and cannonical correlation

library(candisc)

reg3<-lm(as.matrix(dasub1[,-1])~dasub1$x4)

regcanno<-candisc(reg3)

regcanno

#discriminant analysis
library(MASS)

da11<-lda(x2~x6+x7+x8+x9+x10+x11+x12+x13+x14+x15+x16+x17+x18, data= da1, na.action = "na.omit", cv=TRUE)
summary(da11)
da11$prior
da11$counts
da11$means
da11$scaling  # canonical discriminat function
da11$svd

#linear discriminat analysis (classification function)
library(DiscriMiner)
dasub1$x2<-factor(dasub1$x2)

da12<-linDA(dasub1[,-1],dasub1$x2)
da12
summary(da12)
da12$functions  # fisher classification function
da12$confusion
da12$scores
da12$classification
da12$error_rate

discPower(dasub1[,-1],dasub1$x2)  # # discriminating power, wilks lambda