You are on page 1of 40

Multivariate Data Analysis (MNG 744)

Assignment 2
SUBMITTED BY:
Arijit Maji (194270001),
Nilendra Singh Pawar (194278001),
Rahul Sahu(194270003),
VedBeloskar(184278001).

1. Exercise 4.28 page no. 206.

R Code

library(matlib)
library(ggpubr)
library(nortest)
data=read.csv("C:/Users/91907/Desktop/Multivariate
Assignment/Assignment2/Air Quality.csv")

solar_radiation<-data$Solar.radiation...x2.

## Normality check....

hist(solar_radiation)
qqnorm(solar_radiation,pch = 1)
qqline(solar_radiation, col = "steelblue", lwd = 2)
# normality check using correaltion....

len<-nrow(data)

quantile_val<-c()
normal_quantile<- function(x){
for(i in 1:length(x)){
q=(i-.5)/len
quantile_val[i]= qnorm(q)
}
return(quantile_val)
}

Z_value<-normal_quantile(sort(solar_radiation))

r_val<-
cov(sort(solar_radiation),Z_value)/(sd(sort(solar_radiation))*sd(Z_
value))
r_critical<-.9726

R Output

hist(solar_radiation)
qqnorm(solar_radiation,pch = 1)
qqline(solar_radiation, col = "steelblue", lwd = 2)
ad.test(solar_radiation)

Anderson-Darling normality test

data: solar_radiation
A = 0.96228, p-value = 0.01374

Comments:
i. Histogram not showing a bell-shaped curve.
ii. The points on the Q-Q plot nearly fit a straight line and it showing
negatively skewed distribution.
iii. The AD test for normality, p-value is less than .05, reject H0 (the
data is normally distributed) at 5% level of significance.
iv. Hence the variable is not normally distributed.

 Normality Check using correlation.

Z_value<-normal_quantile(sort(solar_radiation))
r_val<-
cov(sort(solar_radiation),Z_value)/(sd(sort(solar_radiation))*sd(Z_value)
)

r_val

[1] 0.9693258

r_critical

[1] 0.9726

Since r_val< r_critical, we cannot reject the Null hypothesis and it is not
normally distribute.

2. Exercise 4.33 Page no 206.

R Code.

q2 <- read.delim("~/phD/SEM 2/Multivariate analysis/assignment 2/stiffness


data.txt")

library("carData")
library("car")

#Normality of x1
hist(q2$x1)
mean(q2$x1)
median(q2$x1)
qqPlot(q2$x1)

q<-data.frame(cbind(sort(q2$x1),seq(1,length(q2$x1),1)))
colnames(q)<-c("x1","z1")
q$z1<-qnorm(((q$z1-.5)/length(q$z1)))
cor(q$z1,q$x1)
r_critial<-0.9652
shapiro.test(q2$x1)

#Normality of x2

hist(q2$x2)
mean(q2$x2)
median(q2$x2)
qqPlot(q2$x2)

q<-data.frame(cbind(sort(q2$x2),seq(1,length(q2$x2),1)))
colnames(q)<-c("x2","z2")
q$z2<-qnorm(((q$z2-.5)/length(q$z2)))
cor(q$z2,q$x2)
r_critial<-0.9652
shapiro.test(q2$x2)

#Normality of x3

hist(q2$x3)
mean(q2$x3)
median(q2$x3)
qqPlot(q2$x3)
q<-data.frame(cbind(sort(q2$x3),seq(1,length(q2$x3),1)))
colnames(q)<-c("x3","z3")
q$z3<-qnorm(((q$z3-.5)/length(q$z3)))
cor(q$z3,q$x3)
r_critial<-0.9652
shapiro.test(q2$x3)

#Normality of x4

hist(q2$x4)
mean(q2$x4)
median(q2$x4)
qqPlot(q2$x4)

q<-data.frame(cbind(sort(q2$x4),seq(1,length(q2$x4),1)))
colnames(q)<-c("x4","z4")
q$z4<-qnorm(((q$z4-.5)/length(q$z4)))
cor(q$z4,q$x4)
r_critial<-0.9652
shapiro.test(q2$x4)

#Bivariate Normality x1 x2
data2<-q2[,c("x1","x2")]
cm<-colMeans(data2)
S<-cov(data2)
d <- apply(data2, MARGIN = 1, function(data2) t(data2-cm) %*% solve(S)
%*% (data2-cm))
plot(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2), sort(d), xlab =
expression(paste(chi[2]^2, "Quantile")), ylab = "Ordered Distances", main =
"Chi-square q-q Plot for (x1,x2)"); abline(a=0, b=1)
cor(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2),sort(d))
#Bivariate Normality x1 x3
data2<-q2[,c("x1","x3")]
cm<-colMeans(data2)
S<-cov(data2)
d <- apply(data2, MARGIN = 1, function(data2) t(data2-cm) %*% solve(S)
%*% (data2-cm))
plot(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2), sort(d), xlab =
expression(paste(chi[2]^2, "Quantile")), ylab = "Ordered Distances", main =
"Chi-square q-q Plot for (x1,x3)"); abline(a=0, b=1)
cor(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2),sort(d))

#Bivariate Normality x1 x4
data2<-q2[,c("x1","x4")]
cm<-colMeans(data2)
S<-cov(data2)
d <- apply(data2, MARGIN = 1, function(data2) t(data2-cm) %*% solve(S)
%*% (data2-cm))
plot(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2), sort(d), xlab =
expression(paste(chi[2]^2, "Quantile")), ylab = "Ordered Distances", main =
"Chi-square q-q Plot for (x1,x4)"); abline(a=0, b=1)
cor(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2),sort(d))

#Bivariate Normality x2 x3
data2<-q2[,c("x2","x3")]
cm<-colMeans(data2)
S<-cov(data2)
d <- apply(data2, MARGIN = 1, function(data2) t(data2-cm) %*% solve(S)
%*% (data2-cm))
plot(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2), sort(d), xlab =
expression(paste(chi[2]^2, "Quantile")), ylab = "Ordered Distances", main =
"Chi-square q-q Plot for (x2,x3)"); abline(a=0, b=1)
cor(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2),sort(d))

#Bivariate Normality x2 x4
data2<-q2[,c("x2","x4")]
cm<-colMeans(data2)
S<-cov(data2)
d <- apply(data2, MARGIN = 1, function(data2) t(data2-cm) %*% solve(S)
%*% (data2-cm))
plot(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2), sort(d), xlab =
expression(paste(chi[2]^2, "Quantile")), ylab = "Ordered Distances", main =
"Chi-square q-q Plot for (x2,x4)"); abline(a=0, b=1)
cor(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2),sort(d))

#Bivariate Normality x3 x4
data2<-q2[,c("x3","x4")]
cm<-colMeans(data2)
S<-cov(data2)
d <- apply(data2, MARGIN = 1, function(data2) t(data2-cm) %*% solve(S)
%*% (data2-cm))
plot(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2), sort(d), xlab =
expression(paste(chi[2]^2, "Quantile")), ylab = "Ordered Distances", main =
"Chi-square q-q Plot for (x3,x4)"); abline(a=0, b=1)
cor(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2),sort(d))

R Output:

> q2 <- read.delim("~/phD/SEM 2/Multivariate analysis/assignment


2/stiffness data.txt")
> library("carData")
> library("car")
>
> #Normality of x1
>
>hist(q2$x1)
> mean(q2$x1)
[1] 1906.1
> median(q2$x1)
[1] 1863
>qqPlot(q2$x1)
[1] 9 17
>
>
> q<-data.frame(cbind(sort(q2$x1),seq(1,length(q2$x1),1)))
>colnames(q)<-c("x1","z1")
> q$z1<-qnorm(((q$z1-.5)/length(q$z1)))
>cor(q$z1,q$x1)
[1] 0.9598642
>r_critial<-0.9652
#As r_value< r_ we can say that the data is not normal, but the
values are very close
>shapiro.test(q2$x1)

Shapiro-Wilk normality test

data: q2$x1
W = 0.93068, p-value = 0.05118

#As p-value is >0.05, Shapiro-Wilk test shows data is normal. The p-


value is barely more than 0.05, so the normality of this data is
borderli

Comments:
v. Histogram appears to roughly fit a bell-shaped curve.
vi. Mean and median are nearby each other.
vii. The points on the Q-Q plot nearly fit a straight line and correlation
between the quantiles is high (>90%).
viii. Based on r-value and Shapiro wilk test this data is borderline
normal

>
>#Normality of x2
>>hist(q2$x2)
> mean(q2$x2)
[1] 1749.533
> median(q2$x2)
[1] 1680
>qqPlot(q2$x2)
[1] 9 17
>
> q<-data.frame(cbind(sort(q2$x2),seq(1,length(q2$x2),1)))
>colnames(q)<-c("x2","z2")
> q$z2<-qnorm(((q$z2-.5)/length(q$z2)))
>cor(q$z2,q$x2)
[1] 0.9503909
>r_critial<-0.9652
#As r_value< r_ we can say that the data is not normal

>shapiro.test(q2$x2)

Shapiro-Wilk normality test

data: q2$x2
W = 0.91274, p-value = 0.01746
#As p-value < 0.05, the data is not normally distributed.
Comments:
i. Visually the data is skewed towards right, and the points on q-q
plot deviate from the straight line at many places.
ii. Based on the r-value and shapirowilk’s test the data is not
normal

>
>
> #Normality of x3
>
>hist(q2$x3)
> mean(q2$x3)
[1] 1509.133
> median(q2$x3)
[1] 1466
>qqPlot(q2$x3)
[1] 9 2
>
> q<-data.frame(cbind(sort(q2$x3),seq(1,length(q2$x3),1)))
>colnames(q)<-c("x3","z3")
> q$z3<-qnorm(((q$z3-.5)/length(q$z3)))
>cor(q$z3,q$x3)
[1] 0.9634104
>r_critial<-0.9652
#As r_value< r_ we can say that the data is not normal, but the r
values are very close.

>shapiro.test(q2$x3)

Shapiro-Wilk normality test

data: q2$x3
W = 0.93258, p-value = 0.05751

#As p-value is >0.05 we can say that the data is normal, but barely
so.
Comments:
i. Histogram shows that the data is positively skewed.
ii. Mean and median are nearby each other.
iii. Based on r-value and Shapiro wilk test this data is borderline
normal

> #Normality of x4
>
>hist(q2$x4)
> mean(q2$x4)
[1] 1724.967
> median(q2$x4)
[1] 1674.5
>qqPlot(q2$x4)
[1] 9 17
>
> q<-data.frame(cbind(sort(q2$x4),seq(1,length(q2$x4),1)))
>colnames(q)<-c("x4","z4")
> q$z4<-qnorm(((q$z4-.5)/length(q$z4)))
>cor(q$z4,q$x4)
[1] 0.980292
> r_<-0.9652
#As r_value> r_ we can say that the data is normal

>shapiro.test(q2$x4)

Shapiro-Wilk normality test

data: q2$x4
W = 0.96127, p-value = 0.3337
#As p-value > 0.05 the data is normally distributed
Comments:
i. Histogram appears to roughly fit a bell-shaped curve.
ii. Mean and median are nearby each other.
iii. Based on the r value and Shapiro-wilk test, we can say that the data
is normally distributed.
SUMMARY: x1 & x3:borderline normal, x4: normal, x2: not
normal

>
>
> #Bivariate Normality x1 x2
> data2<-q2[,c("x1","x2")]
> cm<-colMeans(data2)
> S<-cov(data2)
> d <- apply(data2, MARGIN = 1, function(data2) t(data2-cm) %*%
solve(S) %*% (data2-cm))
> plot(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2), sort(d), xlab =
expression(paste(chi[2]^2, "Quantile")), ylab = "Ordered Distances",
main = "Chi-square q-q Plot for (x1,x2)"); abline(a=0, b=1)
>cor(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2),sort(d))
[1] 0.966762
>

Comments:
i. Correlation between ordered distances between X1 & X2 and chi-
square quantiles = 0.966762
ii. The points on the chi-square plot (except last 4 points) fit a 450
straight line originating from the origin (of the form y=x).
iii. However, as x2 is not normally distributed, we can conclude that
the two variables X1 and X2 do not follow bivariate normal
distribution

>
> #Bivariate Normality x1 x3
> data2<-q2[,c("x1","x3")]
> cm<-colMeans(data2)
> S<-cov(data2)
> d <- apply(data2, MARGIN = 1, function(data2) t(data2-cm) %*%
solve(S) %*% (data2-cm))
> plot(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2), sort(d), xlab =
expression(paste(chi[2]^2, "Quantile")), ylab = "Ordered Distances",
main = "Chi-square q-q Plot for (x1,x3)"); abline(a=0, b=1)
>cor(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2),sort(d))
[1] 0.9824718

Comments:
i. Correlation between ordered distances between X1 & X3 and chi-
square quantiles = 0.9824718
ii. The points on the chi-square plot (except two outliers) fit a 450
straight line originating from the origin (of the form y=x).
iii. The correlation between ordered distances between X1 & X3 and
chi-square quantiles is high (>90%)Hence the two variables X1
and X3 seem to follow bivariate normal distribution.

>
> #Bivariate Normality x1 x4
> data2<-q2[,c("x1","x4")]
> cm<-colMeans(data2)
> S<-cov(data2)
> d <- apply(data2, MARGIN = 1, function(data2) t(data2-cm) %*%
solve(S) %*% (data2-cm))
> plot(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2), sort(d), xlab =
expression(paste(chi[2]^2, "Quantile")), ylab = "Ordered Distances",
main = "Chi-square q-q Plot for (x1,x4)"); abline(a=0, b=1)
>cor(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2),sort(d))
[1] 0.925593

Comments:
i. Correlation between ordered distances between X1 & X4 and chi-
square quantiles = 0.925593
ii. The points on the chi-square plot (except two outliers) do not fit a
450 straight line originating from the origin (of the form y=x). The
curve indicates that the data is positively skewed.
iii. Hence the two variables X1 and X4 not jointly normal distribution.

> #Bivariate Normality x2 x3


> data2<-q2[,c("x2","x3")]
> cm<-colMeans(data2)
> S<-cov(data2)
> d <- apply(data2, MARGIN = 1, function(data2) t(data2-cm) %*%
solve(S) %*% (data2-cm))
> plot(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2), sort(d), xlab =
expression(paste(chi[2]^2, "Quantile")), ylab = "Ordered Distances",
main = "Chi-square q-q Plot for (x2,x3)"); abline(a=0, b=1)
>cor(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2),sort(d))
[1] 0.9440135

Correlation between ordered distances between X2 & X3 and chi-square


quantiles = 0.9440135

Comments:
i. Correlation between ordered distances between X2 & X3 and chi-
square quantiles = 0.9440135
ii. The points on the chi-square plot (except two outliers) do not fit a
450 straight line originating from the origin (of the form y=x).
iii. Furthermore, x2 is not normally distributed, hence the two
variables X2 and X3 do not follow bivariate normal distribution.

>
> #Bivariate Normality x2 x4
> data2<-q2[,c("x2","x4")]
> cm<-colMeans(data2)
> S<-cov(data2)
> d <- apply(data2, MARGIN = 1, function(data2) t(data2-cm) %*%
solve(S) %*% (data2-cm))
> plot(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2), sort(d), xlab =
expression(paste(chi[2]^2, "Quantile")), ylab = "Ordered Distances",
main = "Chi-square q-q Plot for (x2,x4)"); abline(a=0, b=1)
>cor(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2),sort(d))
[1] 0.9219479

Comments:
i. Correlation between ordered distances between X2 & X4 and chi-
square quantiles = 0.9219479
ii. The points on the chi-square plot (except three outliers) do not fit a
450 straight line originating from the origin (of the form y=x).
iii. X2 is not normally distributed, so the two variables X2 and X4 do
not follow bivariate normal distribution.

>
> #Bivariate Normality x3 x4
> data2<-q2[,c("x3","x4")]
> cm<-colMeans(data2)
> S<-cov(data2)
> d <- apply(data2, MARGIN = 1, function(data2) t(data2-cm) %*%
solve(S) %*% (data2-cm))
> plot(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2), sort(d), xlab =
expression(paste(chi[2]^2, "Quantile")), ylab = "Ordered Distances",
main = "Chi-square q-q Plot for (x3,x4)"); abline(a=0, b=1)
>cor(qchisq((1:nrow(data2)-1/2)/nrow(data2),df=2),sort(d))
[1] 0.9839645

Comments:
i. Correlation between ordered distances between X3 & X4 and chi-
square quantiles = 0.9839645
ii. The points on the chi-square plot (except two outliers) fit a 450
straight line originating from the origin (of the form y=x).

iii. Hence the two variables X3 and X4 follow bivariate normal


distribution.

3. Exercise 4.39 Page no 207.


R Code

data1=read.csv("C:/Users/91907/Desktop/MultivariateAssignment/Psyc
hological_Data.csv")

# Marginal normality check code

hist(data1$Indep)
qqnorm(data1$Indep,pch = 1)
qqline(data1$Indep, col = "steelblue", lwd = 2)
shapiro.test(data1$Indep)

hist(data1$Supp)
qqnorm(data1$Supp,pch = 1)
qqline(data1$Supp, col = "steelblue", lwd = 2)
shapiro.test(data1$Supp)

hist(data1$Benev)
qqnorm(data1$Benev,pch = 1)
qqline(data1$Benev, col = "steelblue", lwd = 2)
shapiro.test(data1$Benev)

hist(data1$Conform)
qqnorm(data1$Conform,pch = 1)
qqline(data1$Conform, col = "steelblue", lwd = 2)
shapiro.test(data1$Conform)

hist(data1$Leader)
qqnorm(data1$Leader,pch = 1)
qqline(data1$Leader, col = "steelblue", lwd = 2)
shapiro.test(data1$Leader)

## Joinnly multivariate normality checks code.

# distance calculation code.

mean_vec<-colMeans(data1)
X_Minus_mu<-t(t(data1)-mean_vec)

var_cov<- cov(data1)
inverse_mat<- inv(var_cov)

dd<- diag(X_Minus_mu %*% var_cov%*% t(X_Minus_mu))

# quantile calculation code.

sort_dd<-sort(dd)
len<-length(dd)
DOF<-ncol(data1)
chi_val<-c()
chi_sq_value<- function(x,dof){
for(i in 1:length(x)){
q=(i-.5)/len
chi_val[i]= qchisq(q,dof)
}
return(chi_val)
}

q_value<-chi_sq_value(sort_dd,DOF)

chi_value<-qchisq(.95,DOF)
percentage<- (sum(q_value<=qchisq(.95,DOF)))/len

hist(q_value)
plot(q_value,sort_dd)

comment<- if(percentage>=0.95){
print("Multivariate Normal Distribution")
}

# Power transform code.

df=data1
a=powerTransform(data1$Supp, family="bcPower")
a
pow<-1.396262
df$Supp = ((df$Supp)**pow -1)/pow

hist(df$Supp)
qqnorm(df$Supp,pch = 1)
qqline(df$Supp, col = "steelblue", lwd = 2)
shapiro.test(df$Supp)

b=powerTransform(data1$Leader, family="bcPower")
b
pow<-0.3815471
df$Leader = ((df$Leader)**pow -1)/pow

hist(df$Leader)
qqnorm(df$Leader,pch = 1)
qqline(df$Leader, col = "steelblue", lwd = 2)
shapiro.test(df$Leader)
## Joinnly multivariate normality check after power transform.

mean_vec<-colMeans(df)

X_Minus_mu<- t(t(df)-mean_vec)

var_cov<- cov(df)
inverse_mat<- inv(var_cov)
dd<- diag(X_Minus_mu %*% var_cov%*% t(X_Minus_mu))
sort_dd<-sort(dd)

len<-length(dd)
DOF<-ncol(data1)
chi_val<-c()
chi_sq_value<- function(x,dof){
for(i in 1:length(x)){
q=(i-.5)/len
chi_val[i]= qchisq(q,dof)
}
return(chi_val)
}

q_value<-chi_sq_value(sort_dd,DOF)

chi_value<-qchisq(.95,DOF)

percentage_1<- (sum(q_value<=qchisq(.95,DOF)))/len

hist(q_value)
plot(q_value,sort_dd)
comment<- if(percentage_1>=0.95){
print("Multivariate Normal Distribution")}

R Output

Shapiro-Wilk normality test

data: data1$Indep
W = 0.97486, p-value = 0.01617

Comments:
i. Histogram is showing a bell-shaped curve, showing right skewed.
ii. The points on the Q-Q plot not fit a strong straight line and it
showing negatively skewed distribution.
iii. The Shapiro-Wilk test for normality, p-value is less than .05, reject
H0 (the data is normally distributed) at 5% level of significance.
iv. Hence the variable is not normally distributed.
Shapiro-Wilk normality test

data: data1$Supp
W = 0.97788, p-value = 0.03198

Comments:
v. Histogram is showing a bell-shaped curve , most of the values are
right side concentrated,.
vi. The points on the Q-Q plot not fit a strong straight line and it
showing slightly positivelye skewed distribution.
vii. The Shapiro-Wilk test for normality, p-value is less than .05, reject
H0 (the data is normally distributed) at 5% level of significance.
viii. Hence the variable is not normally distributed.
Histogram of Benev
40
30
Frequency

20
10
0

0 5 10 15 20 25 30

Benev

Normal Q-Q plot of Benev


30
20 25
Sample Quantiles

15
10
5

-2 -1 0 1 2

Theoretical Quantiles

Shapiro-Wilk normality test

data: data1$Benev
W = 0.98387, p-value = 0.1269
i. Histogram is showing a tight bell-shaped curve.
ii. The points on the Q-Q plot showing a fit strong straight line and it
showing slightly positively skewed distribution.
iii. The Shapiro-Wilk test for normality, p-value is greater than .05,
not reject H0 (the data is normally distributed) at 5% level of
significance.
iv. Hence the variable is normally distributed.
Shapiro-Wilk normality test

data: data1$Conform
W = 0.98461, p-value = 0.15

i. Histogram is showing not a tight bell-shaped curve.


v. The points on the Q-Q plot showing a fit strong straight line and
only lower and upper values are not fitting the straight line.
vi. The Shapiro-Wilk test for normality, p-value is greater than .05,
not reject H0 (the data is normally distributed) at 5% level of
significance.
vii. Hence the variable is normally distributed.
Shapiro-Wilk normality test

data: data1$Leader
W = 0.96256, p-value = 0.001196
i. Histogram is showing not a bell-shaped curve.
ii. The points on the Q-Q plot not showing a fit strong straight line
and both side distribution has extreme values.
iii. The Shapiro-Wilk test for normality, p-value is less than .05,
reject H0 (the data is normally distributed) at 5% level of
significance.
iv. Hence the variable is not normally distributed.

 Jointly multivariate normality check.


percentage<- (sum(q_value<=qchisq(.95,DOF)))/len
0.9538462
i. Histogram is showing a rightly skewed distribution, similar tp chi-
sq distribution.
ii. The points on the Q-Q plot showing a fit strong straight line.
iii. More than 95% observations lie below the critical value
(14.06714).
v. Hence the variables are jointly normally distributed.

 Power Transformation.
Shapiro-Wilk normality test

data: df$Supp

W = 0.98305, p-value = 0.105


Shapiro-Wilk normality test

data: df$Leader
W = 0.99258, p-value = 0.7276
percentage_1<- (sum(q_value<=qchisq(.95,DOF)))/len
[1] 0.9538462
More than 95% observations lie below the critical value (14.06714).
Q-Q plot shows a straight line.
Joint distribution is normally distributed.

After and before power transformation of variables, there is no such


significant difference, only fluctuation of Q-Q plot after transforming is
less compare before transforming.

You might also like