Review on Chapter 5 Part
III
In Chapter 5 Part III (p.48-p.63), we have learnt:
1. See Example 9 in Chapter 5 for ( F -test + t -test)
Part 1: Check if σ 21 and σ 22 are equal
H 0 :σ 1=σ 2(Case B.1) v.s. H 1 : σ 21 ≠ σ 22 (Case B.2)
2 2
Part 2: Perform t-test assuming σ 21=σ 22, i.e., Perform
pooled variance t -test, if Case B.1 is not rejected in
Part 1.
2. One Way ANOVA - Test on Equality of k population
means
Step 1:
H 0 : μ1=μ2=…=μ k
H 1 : Not all means are equal
Step 2:
Sample size of i-th sample¿ ni
Sample mean of i-th sample¿ x i
Sample variance of i-th sample ¿ s2i
k
Total sample size n=∑ ni=n1+ n2 +…+n k
i=1
k
Grand mean: ∑ ni x i
1
x= i=1
= (n1 x 1+ n2 x 2 +…+n k x k )
n n
Sum of squares for treatments SS T r:
k
SS T r=∑ n i ( x i−x )
2
i=1
Sum of squares for errors SSE:
k
SSE=∑ (ni −1)s 2i
i=1
Then the One-way ANOVA Table is as follows:
Source Degrees Sums of Mean F-
of of Squares Squares statistic
Variatio freedom
n
Treatme k −1 SS T r SS T r MS T r
MS T r = F=
nts k−1 MSE
Error n−k SSE SSE
MSE=
n−k
Total n−1 SS T r +SSE
The test statistic for Step 2 is F= MS T r
MSE
Step 3:
The Rejection Region Method
The rejection region can be found using F
Table with significance level α and degree of
freedoms:
Numerator d . f .=k−1
Denominator d . f .=n−k
Do a right-tailed test with α =¿right-tail area.
P.6
Simple Linear Regression Model
-Interest: We want to study the relationship
between Y (dependent variable) and X
(independent variable).
-Mathematical Model used: Simple Linear
Regression Model
Y = β0 + β 1 X +ϵ
where ϵ is error variable (account for the
deviation).
-Sample data set: contains n data points
{( x1 , y 1 ) , ( x 2 , y 2) , ( x 3 , y 3 ) , … , ( x n−1 , y n−1 ) , ( x n , y n ) }
P.7
Step 1 (Scatter Plot):
Draw a scatter plot to have a first look of whether
Y is linearly related to X . There are four possible
scenarios:
Scenario 1:
It means that X and Y have perfect linear
relation.
Scenario 2:
It means that X and Y have strong linear
relation.
Scenario 3:
It means that X and Y have weak linear
relation.
Scenario 4:
It means that X and Y have no linear relation.
P.13
Step 3 (Prediction or Estimation):
Make prediction or estimation using the
regression line.
Let say you have a new data with known value x ¿,
you want to estimate the value of y ¿. Then the
estimated value of y ¿ will be
¿ ¿
^y =b0 + b1 x
See Example 1 (c).
P.19
Strength of linear relationship - Coefficient of
Correlation
The coefficient of correlation measures the
strength of linear relationship between X and Y .
Sample version: r (sample statistic)
Population version: ρ (population parameter)
They are between -1 and 1, i.e. −1 ≤r ≤ 1 and −1 ≤ ρ≤ 1.
Three extreme cases for samples:
Perfect, positive linear relation (r =1¿
Perfect, negative linear relation (r =−1¿
No linear pattern (r =0 ¿
Use calculator to find r . See the file Linear
Regression - Calculator [Link] for instructions.
P.22
Strength of linear relationship - Coefficient of
Determination
2 2
R = ( coefficient of correlation )
It measures the percentage of variation in the
dependent variable that can be explained by the
regression line. (how much Y is
controlled/affected by X )
P.25
Hypothesis Testing – Testing of Linear Relationship
Step 1: (One of the following based on the given
claim)
H 0 : ρ=0 v.s. H 1 : ρ>0
H 0 : ρ=0 v.s. H 1 : ρ<0
H 0 : ρ=0 v.s. H 1 : ρ≠ 0
Step 2:
√
Test statistic t=r n−22
1−r
Step 3:
The Rejection Region Method
The rejection region can be found based on 3
things:
(a) t -Distribution with degree of freedom n−2
(b) H 1 (which tail(s) need to be considered)
(c) significance level α
P.27
Example 1 (Revisited) – Testing of Linear
Relationship
Recall that the sample coefficient of correlation was
found to be
r =0.7.
Test if X and Y for the population are actually
linearly related at α =5 %.
Solution:
Claim: X and Y for the population are actually
linearly related.
Claim: ρ ≠ 0
Step 1:
H 0 : ρ=0 v.s. H 1 : ρ≠ 0 (Claim)
Step 2:
Sample size n=6
Sample coefficient of correlation r =0.7
√
Test statistic t=r n−22 =0.7
1−r √ 6−2
1−0.7 2
=1.9604
Degree of freedom ¿ n−2=4
Step 3:
Significance level α =0.05
For Rejection Region Method:
Based on the H 1: ¿ in this test¿
Since t 0.025 ,4 =2.7764 and t 0.975 ,4 =−t0.025 , 4=−2.7764
The Rejection Region is t ←2.7764 or t >2.7764.
Conclusion:
Test statistic 1.9604 does not fall in the
rejection region. (−2.7764 <1.9604<2.7764 ¿
We do not reject H 0 at α =5 % .
There is not enough evidence to infer that the
claim ( H 1) is true.
So there is not enough evidence to infer that
X is linearly related to Y .