F24 QMS 210

Lecture 8-2
F-Test

Dr Boža Tasić

October 29, 2024

F Test for the Ratio of Two Variances

I We often need to determine if the two independent

populations have the same amount of variability. Testing
variances we can detect the differences in the amount of
variability.
I One instance when testing variances is very important is to
determine whether to use
I The pooled-variance t test ( which assumes equal variances) or
I the separate-variances t test (which does not assume equal
variances) when comparing two means.
F Test for the Ratio of Two Variances

I Assumptions:
I We have two populations that are normally distributed!
Population 1 and population 2. Data collected in both
samples are numerical.
I We take a random sample of size n1 from population 1 and a
random sample of size n2 from population 2. Sample selected
from one population has no effect on the sample selected from
the other population, i.e., samples are independent.
I The variance of the sample 1 (from population 1) is denoted
by S12 . The variance of the sample 2 (from population 2) is
denoted by S22
F Test for the Ratio of Two Variances

I Our goal: For a given level of significance to test hypothesis

for the difference (<, 6=, >) between two population variances.
I To test
H0 : σ12 = σ22 , H1 : σ12 6= σ22
we use Two-tail Test.
I To test
H0 : σ12 ≤ σ22 , H1 : σ12 > σ22
we use Upper-tail Test (Right- tail Test).
I To test
H0 : σ12 ≥ σ22 , H1 : σ12 < σ22
we use Lower-tail Test (Left-tail Test).
F Test for the Ratio of Two Variances

I To test the difference (<, 6=, >) of two variances, we define

the test statistic FSTAT by

S12
FSTAT =
S22

where
I S12 and S22 are variances of the samples taken from
populations 1 and 2 of sizes n1 and n2 respectively.
I n1 − 1 = degrees of freedom from sample 1 (the numerator
degrees of freedom)
I n2 − 1 = degrees of freedom from sample 2 (the denominator
degrees of freedom)
F Test for the Ratio of Two Variances
I This test is called F -Test for the Ratio of Two Variances.
The test statistic
S2
FSTAT = 12
S2
follows an F distribution with n1 − 1 and n2 − 1 degrees of
freedom.

I
I The F distribution is a family of curves based on the degrees
of freedom of the variance of the numerator and the degrees
of freedom of the variance of the denominator.
F Test for the Ratio of Two Variances

I Notice that the F distribution curve is not symmetric (it is not

a bell curve!!)
I The values of F cannot be negative, because variances are
always positive or zero.
I The distribution is positively skewed.
I The critical values of the F distribution depend on the degrees
of freedom in the two samples and they can be found using
the CASIO calculator.
Finding the Rejection Region for F Test
I For a two tail test, i.e. if we are testing
H0 : σ12 = σ22 , H1 : σ12 6= σ22
we will have the following rejection region

I
I We use FU = F α2 ,n1 −1,n2 −1 as the upper critical value.
I The lower critical value
1
FL = F1− α2 ,n1 −1,n2 −1 =
F α2 ,n2 −1,n1 −1
Finding the Rejection Region for F Test
I For a lower tail test, i.e. if we are testing
H0 : σ12 ≥ σ22 , H1 : σ12 < σ22
we will have the following rejection region

I
I We use the lower critical value
1
FL = F1−α,n1 −1,n2 −1 =
Fα,n2 −1,n1 −1
Finding the Rejection Region for F Test
I For an upper tail test, i.e. if we are testing

H0 : σ12 ≤ σ22 , H1 : σ12 > σ22

we will have the following rejection region

I We use FU = Fα,n1 −1,n2 −1 as the upper critical value.

F Test for the Ratio of Two Variances - Example 1

I You are a financial analyst for a brokerage firm. Is there a

difference in the variances between the stocks listed on the
NYSE (New York Stock Exchange) and NASDAQ (National
Association of Securities Dealers Automated Quotations) at
α = 0.05 level of significance? You collect the following data:
I
F Test for the Ratio of Two Variances - Example 1
1. State the appropriate null and alternative hypothesis.
I Let’s denote by σ12 the variance of NYSE stocks and by σ22 the
variance of NASDAQ stocks. We are testing if there is a
difference, so

H0 : σ12 = σ22 , H1 : σ12 6= σ22

I H0 : σ12 = σ22 means there is no difference between variances.

I H1 : σ12 6= σ22 means there is a difference between variances.
2. Specify the desired level of significance and the sample sizes.
I We have α = 0.05. The NYSE sample size n1 = 21 and the
NASDAQ sample size n2 = 25. So,
I The numerator degrees of freedom df1 = 21 − 1 = 20 and
I The denominator degrees of freedom df2 = 25 − 1 = 24
3. Determine the appropriate technique
I This is an F Test for the Ratio of Two Variances
F Test for the Ratio of Two Variances - Example 1
4. Determine the critical values
I The upper critical value FU = F α ,n −1,n −1 of the F
2 1 2
distribution with α = 0.05 and n1 − 1 = df1 = 20 and
n2 − 1 = df2 = 24 can be found using the calculator.
i. Select STAT, F5 (DISTR), F4(F) and then F3(InvF). Now in
the Inverse F menu select the following options:
ii. Data: F2 (Variable)
iii. Area: 0.025 EXE
iv. n:df : 20 EXE
v. d:df : 24 EXE
vi. Save Res: If you do not want to save results in a list press F1
(None). If you want to save results in a list press F2.
vii. Execute: key EXE or F1(Calc).
I You will get the following summary:
I You will get xInv = 2.32727144. So, the critical value rounded
to two decimal places is FU = F0.025,20,24 = 2.33.
F Test for the Ratio of Two Variances - Example 1

I To find the lower critical value use

1
FL = F1− α2 ,n1 −1,n2 −1 =
F α
,n −1,n1 −1
2 2

and the calculator procedure described for the upper critical

value.
I In our case

FL = F1− α2 ,n1 −1,n2 −1 = F0.975,20,24 = 0.4153580

or rounded to two decimal places FL = 0.42.

F Test for the Ratio of Two Variances - Example 1

5. Compute the test statistic

I We have:
I S12 = 1.302 = 1.69,
I S22 = 1.162 = 1.3456
I Then the test statistic is
S12 1.69
FSTAT = = = 1.256
S22 1.3456
F Test for the Ratio of Two Variances - Example 1
6. Is the test statistic in the rejection region? The decision rule
for the two-tail test is
Reject H0 if FSTAT > 2.33 or FSTAT < 0.42
I

6. (continued) Reach a decision and interpret the result

I Since FL = 0.42 < FSTAT = 1.256 < 2.33 = FU , we do not
reject the null hypothesis H0 at α = 0.05 and conclude that
there is insufficient evidence of a difference in variances.
F Test for the Difference Between Two Variances -
Example 1- Using CASIO Calculator
1. Select STAT, F3 (TEST), F4(F). Now select in the
2-Sample FTest the following options:
2. Data: F2 (Variable)
3. σ1 : F1 EXE (6= σ2 )
4. sx1: 1.30 EXE
5. n1 : 21 EXE
6. sx2: 1.16 EXE
7. n2 : 25 EXE
8. Save Res: If you do not want to save results in a list press F1
(None). If you want to save results in a list press F2 and type
the number of list where you want to save results.
9. Execute: key EXE or F1(Calc).
I You will get the following summary:
F Test for the Difference Between Two Variances -
Example 1- Using CASIO Calculator

I 2-Sample FTest
I σ1 6= σ2
I F = 1.2559453
I p = 0.58884619
I sx1 = 1.3
I sx2 = 1.16
I n1 = 21
I n2 = 25
I Statistical Decision: Since p-value = 0.589 > α = 0.05, we
do not reject the null hypothesis H0 at α = 0.05.
I Conclusion: The evidence does not indicate that there is a
difference in variances at α = 0.05.
F Test for the Difference Between Two Variances -
Example 2- Using CASIO Calculator
I We wish to determine if there is a difference in the breaking
distances for two types of tires. Assume that the breaking
distances for each type of tire are normally distributed. Based
on the data for the sample of tires shown, at 5% level of
significance, should we conclude that there is a difference in
the variances between Tire A and Tire B?
Breaking Distance (meters)
Tire A Tire B
83 75
79 84
I 82 76
84 83
80 85
81 78
83
F Test for the Difference Between Two Variances -
Example 2- Using CASIO Calculator
1. Select STAT, F3 (TEST), F4(F). Now select in the
2-Sample FTest the following options:
2. Data: F1 (List)
3. σ1 : F1 EXE (6= σ2 )
4. List(1): List1 EXE
5. List(2): List2 EXE
6. Freq(1): 1 EXE
7. Freq(2): 1 EXE
8. Save Res: If you do not want to save results in a list press F1
(None). If you want to save results in a list press F2 and type
the number of list where you want to save results.
10. Execute: key EXE or F1(Calc).
I You will get the following summary:
F Test for the Difference Between Two Variances -
Example 2- Using CASIO Calculator

I 2-Sample FTest
I σ1 6= σ2
I F = 0.20646067
I p = 0.10432125
I x 1 = 81.5
I x 2 = 80.5714286
I sx1 = 1.87082869
I sx2 = 4.11732692
I n1 = 6
I n2 = 7
F Test for the Difference Between Two Variances -
Example 2- Using CASIO Calculator

I Statistical Decision: Since p-value = 0.10 > α = 0.05, we

do not reject the null hypothesis H0 at α = 0.05.
I Conclusion: The evidence does not indicate that the
variances of the stopping distances are different for the two
types of tires.
Pooled-Variance t Test for µ1 − µ2 - Example 2

I We wish to determine if there is a difference in the breaking

distances for two types of tires. Assume that the breaking
distances for each type of tire are normally distributed. Based
on the data for the sample of tires shown, at 5% level of
significance, should we conclude that there is a difference in
the mean breaking distance?
Breaking Distance (meters)
Tire A Tire B
83 75
79 84
I 82 76
84 83
80 85
81 78
83
Pooled-Variance t Test for µ1 − µ2 - Example 2
1. State the appropriate null and alternative hypothesis.
I Let’s denote by µ1 the mean breaking distance for Tire A and
by µ2 the mean breaking distance for Tire B. We are testing if
there is a difference, so

H0 : µ 1 = µ 2 , H1 : µ1 6= µ2

I H0 : µ1 = µ2 means there is no difference

I H1 : µ1 6= µ2 means there is difference
2. Specify the desired level of significance and the sample sizes.
I We have α = 0.05 and n1 = 6, n2 = 7.
3. Determine the appropriate technique
I This is a pooled-variance t test because in the previous
example we found that the evidence does not indicate that the
variances of the stopping distances are different for the two
types of tires.
Pooled-Variance t Test for µ1 − µ2 - Example 2 - Using
CASIO Calculator
1. Select STAT, F3 (TEST), F2(t) and then F2(2-S). Now
select in the 2-Sample tTest the following options:
2. Data: F1 (List)
3. µ1 : F1 EXE (6= µ2 )
4. List(1): List1 EXE
5. List(2): List2 EXE
6. Freq(1): 1 EXE
7. Freq(2): 1 EXE
8. Pooled: F1 EXE (On). If you select F1(On), you are using
the t test pooled-variance procedure. If you select F2 (Off),
you are using the t test separate-variance procedure.
9. Save Res: If you do not want to save results in a list press F1
(None). If you want to save results in a list press F2 and type
the number of list where you want to save results.
10. Execute: key EXE or F1(Calc).
I You will get the following summary:
Pooled-Variance t Test for µ1 − µ2 - Example 2 - Using
CASIO Calculator
I 2-Sample tTest
I µ1 6= µ2
I t = 0.50699126
I p = 0.62216994
I df = 11
I x 1 = 81.5
I x 2 = 80.6
I sx1 = 1.87
I sx2 = 4.11
I sp = 3.29
I n1 = 6
I n2 = 7
Pooled-Variance t Test for µ1 − µ2 - Example 2 - Using
CASIO Calculator

I Statistical Decision: Since p-value = 0.622 > α = 0.05, we

do not reject the null hypothesis H0 at α = 0.05.
I Conclusion: The evidence does not indicate that the average
stopping distance is different for the two types of tires.
Important Comments for Assignment 8 on MyLab Stats
I When testing

H0 : σ12 = σ22 , H1 : σ12 6= σ22

the larger variance should always be placed in the numerator,

S2
i.e., the test statistic FSTAT = 12 where S12 > S22 .
S2
I If data sets are given as lists, in order to get correct test
statistic find first variances of both lists and see which one has
a larger variance. You will input the list with the larger
variance as List 1 (numerator) and the the list with the
smaller variance as List 2 (denominator). Do this even if your
data sets are labeled in a question as sample 1 and sample 2!
It does not mean that sample 1 has larger variance than
sample 2, you have to check it first. For example, if sample 1
has a smaller variance than sample 2 you will not enter sample
1 as List 1!
Important Comments for Assignment 8 on MyLab Stats
I In MyLab Stats F-test is forced into a right tail test by placing
the sample with the large variance in the numerator and the
smaller variance in the denominator. Remember that it does
not matter which sample has the larger sample size, only
which sample has the larger variance!
I The numerator degrees of freedom will be the degrees of
freedom for the sample with the larger variance (it is in the
numerator of FSTAT ) and the denominator degrees of freedom
will be the degrees of freedom for the sample with the smaller
variance (it is in the denominator of FSTAT ).
I If you are conducting a two-tail test, you still have to divide α
by 2, but you only look up and compare the right critical
value.
I In the template for F-test on CASIO fx-9750GII calculator,
you input standard deviations S1 , S2 ! If variances S12 , S22 are
given instead do not forget to take a square root of them
when inputing them in the calculator.