2. Varsha Varde 2
ā¢ Contents:
ā¢ 1. Introduction
ā¢ 2. Studentās t distribution
ā¢ 3. Small-sample inferences about a population
mean
ā¢ 4. Small-sample inferences about the difference
between two means: Independent Samples
ā¢ 5. Small-sample inferences about the difference
between two means: Paired Samples
ā¢ 6. Inferences about a population variance
ā¢ 7. Comparing two population variances
3. Introduction
ā¢ When the sample size is small we only
deal with normal populations.
ā¢ For non-normal (e.g. binomial) populations
different techniques are necessary
Varsha Varde 3
4. Summary of Test Statistics to be Used in aSummary of Test Statistics to be Used in a
Hypothesis Test about a Population MeanHypothesis Test about a Population Mean
nn >> 30 ?30 ?
ĻĻ known ?known ?
Popul.Popul.
approx.approx.
normalnormal
??
ĻĻ known ?known ?
UseUse ss toto
estimateestimate ĻĻ
UseUse ss toto
EstimateEstimate ĻĻ
IncreaseIncrease nn
toto >> 3030/
x
z
n
Āµ
Ļ
ā
=
/
x
z
s n
Āµā
=
/
x
z
n
Āµ
Ļ
ā
=
/
x
t
s n
Āµā
=
YesYes
YesYes
YesYes
YesYes
NoNo
NoNo
NoNo
NoNo
5. Studentās t Distribution
ā¢ For small samples (n < 30) from normal
populations, we have
ā¢ z =xĀÆ - Āµ/Ļ/ān
ā¢ If Ļ is unknown, we use s instead; but we
no more have a Z distribution
ā¢ Assumptions.
ā¢ 1. Sampled population is normal
ā¢ 2. Small random sample (n < 30)
ā¢ 3. Ļ is unknown
ā¢ t =xĀÆ - Āµ/s/ān
Varsha Varde 5
6. Properties of the t Distribution:
ā¢ (i) It has n - 1 degrees of freedom (df)
ā¢ (ii) Like the normal distribution it has a symmetric
mound-shaped probability distribution
ā¢ (iii) More variable (flat) than the normal distribution
ā¢ (iv) The distribution depends on the degrees of freedom.
Moreover, as n becomes larger, t converges to Z.
ā¢ (v) Critical values (tail probabilities) are obtained from
the t table
ā¢ Examples.
ā¢ (i) Find t0.05,5 = 2.015
ā¢ (ii) Find t0.005,8 = 3.355
ā¢ (iii) Find t0.025,26 = 2.056
Varsha Varde 6
7. Small-Sample Inferences About a Population Mean
ā¢ Parameter of interest: Āµ
ā¢ Sample data: n, xĀÆ, s
ā¢ Other information: Āµ0= target value ;Ī±
ā¢ Point estimator: xĀÆ
ā¢ Estimator mean: ĀµxĀÆ = Āµ
ā¢ Estimated standard errorĻxĀÆ = s/ān
ā¢ Confidence Interval for Āµ:
ā¢ x Ā± tĪ±/2 ,n-1 (s/ān )
Varsha Varde 7
8. Studentās t test
ā¢ Test: H0 : Āµ = Āµ0
ā¢ Ha : 1) Āµ > Āµ0; 2) Āµ < Āµ0; 3) Āµ = Āµ0.
ā¢ Critical value: either tĪ±,n-1 or tĪ±/2 ,n-1
ā¢ T.S. : t = xĀÆ-Āµ0/s/ān
ā¢ RR:1) Reject H0 if t > tĪ±,n-1
ā¢ 2) Reject H0 if t < -tĪ±,n-1
ā¢ 3) Reject H0 if t > tĪ±/2 ,n-1 or t < -tĪ±/2,n-1
ā¢ Decision: 1) if observed value is in RR: āReject H0ā
ā¢ 2) if observed value is not in RR: āDo not reject H0ā
ā¢ Conclusion: At 100Ī±% significance level there is (in)sufficient
statistical evidence to āfavor Haā .
ā¢ Assumptions.1. Small sample (n < 30)
ā¢ 2. Sample is randomly selected
ā¢ 3. Normal population
ā¢ 4. Unknown variance
Varsha Varde 8
9. Example
ā¢ Example: It is claimed that weight loss in a new diet
program is at least 20 pounds during the first month.
Formulate &Test the appropriate hypothesis
ā¢ Sample data: n = 25, xĀÆ =19.3, s2
= 25, Āµ0 = 20, Ī±=0.05
ā¢ Critical value: t0.05,24 = -1.711
ā¢ H0 : Āµ ā„ 20 (Āµ is greater than or equal to 20 )
ā¢ Ha : Āµ < 20,
ā¢ T.S.:t =(xĀÆ - Āµ0 )/s/ān=(19.3 ā 20)/5/ā25 =-0.7
ā¢ RR: Reject H0 if t <-1.711
ā¢ Decision: Reject H0
ā¢ Conclusion: At 5% significance level there is insufficient
statistical evidence to conclude that weight loss in a new
diet program exceeds 20 pounds per first month.
Varsha Varde 9
10. Two Means: Independent Samples
ā¢ Parameter of interest: Āµ1 - Āµ2
ā¢ Sample data:Sample 1: n1, x1, s1 ;Sample 2: n2, x2, s2
ā¢ Other information: D0= target value ; Ī±
ā¢ Point estimator: XĀÆ1 ā XĀÆ2
ā¢ Estimator mean: Āµ XĀÆ1 ā XĀÆ2 = Āµ1 - Āµ2
ā¢ Assumptions.
ā¢ 1. Normal populations 2. Small samples ( n1 < 30; n2 < 30)
ā¢ 3. Samples are randomly selected4. Samples are independent
ā¢ 5. Variances are equal with common variance
ā¢ Ļ2
= Ļ2
1 = Ļ2
2
ā¢ Pooled estimator for Ļ.
ā¢ s = ā[(n1 - 1)s2
1 + (n2 - 1)s2
2
]/(n1 + n2 ā 2)
ā¢ Estimator standard error:
ā¢ ĻXĀÆ1 ā XĀÆ2 = Ļ ā (1/n1+1/n2)
ā¢ Confidence Interval:
ā¢ (xĀÆ1 - xĀÆ2) Ā± (tĪ±/2 ,n1+n2-2)(s ā1/n1+1/n2)
Varsha Varde 10
11. Two Means: Independent Samples
ā¢ Test:H0 : Āµ1 - Āµ2 = D0
ā¢ Ha : 1)Āµ1 - Āµ2 > D0 or
2) Āµ1 - Āµ2 < D0 or
3) Āµ1 - Āµ2 = D0
ā¢ T.S. :t =(xĀÆ1 - xĀÆ2) - D0/s ā1/n1+ 1/n2
ā¢ RR: 1) Reject H0 if t > tĪ±,n1+n2-2
2) Reject H0 if t < - tĪ±,n1+n2-2
3) Reject H0 if t > tĪ±/2,n1+n2-2 or t < - tĪ±/2,n1+n2-2
12. Example.(Comparison of two weight loss programs)
ā¢ Refer to the weight loss example. Test the hypothesis that weight
loss in a new diet program is different from that of an old program.
We are told that that the observed value of Test Statistics is 2.2 and
we know that
ā¢ Sample 1 : n1 = 7 ; Sample 2 : n2 = 8 ; Ī±= 0.05
ā¢ Solution.
ā¢ H0 : Āµ1 - Āµ2 = 0
ā¢ Ha : Āµ1 - Āµ2 ā 0
ā¢ T.S. :t =(x1 - x2) ā 0/s ā1/n1+ 1/n2= 2.2
ā¢ Critical value: t.025,13 = 2.16
ā¢ RR: Reject H0 if t > 2.160 or t < -2.160
ā¢ Decision: Reject H0
ā¢ Conclusion: At 5% significance level there is sufficient
statistical evidence to conclude that weight loss in the
two diet programs are differentVarsha Varde 12
13. Small-Sample Inferences About the Difference
Between
Two Means: Paired Samples
ā¢ Parameter of interest: Āµ1 - Āµ2 = Āµd
ā¢ Sample of paired differences data:
ā¢ Sample : n = number of pairs, dĀÆ = sample mean, sd
ā¢ Other information: D0= target value, Ī±
ā¢ Point estimator: dĀÆ
ā¢ Estimator mean: ĀµdĀÆ = Āµd
ā¢ Assumptions.
ā¢ 1. Normal populations
ā¢ 2. Small samples ( n1 < 30; n2 < 30)
ā¢ 3. Samples are randomly selected
ā¢ 4. Samples are paired (not independent)
Varsha Varde 13
14. ā¢ Sample standard deviation of the sample
of n paired differences
n
ā¢ sd = ā Ī£ (di - ĀÆd)2
/ n-1
i=1
ā¢ Estimator standard error: Ļd = sd/ān
ā¢ Confidence Interval.
ā¢ ĀÆd Ā± tĪ±/2,n-1sd/ān
Varsha Varde 14
15. Test.
ā¢ H0 : Āµ1 - Āµ2 = D0 (equivalently, Āµd = D0)
ā¢ Ha : 1) Āµ1 - Āµ2 = Āµd > D0;
ā¢ 2) Āµ1 - Āµ2 = Āµd < D0;
ā¢ 3) Āµ1 - Āµ2 = Āµd = D0,
ā¢ T.S. :t =ĀÆd - D0/sd/ān
ā¢ RR:
ā¢ 1) Reject H0 if t > tĪ±, n-1
ā¢ 2) Reject H0 if t < -tĪ±,n-1
ā¢ 3) Reject H0 if t > tĪ±/2,n-1 or t < -tĪ±/2,n-1
Varsha Varde 15
16. Example.
ā¢ A manufacturer wishes to compare wearing qualities of
two different types of tires, A and B. For the comparison
a tire of type A and one of type B are randomly assigned
and mounted on the rear wheels of each of five
automobiles. The automobiles are then operated for a
specified number of miles, and the amount of wear is
recorded for each tire. These measurements are
tabulated below.
ā¢ Automobile Tire A Tire B
ā¢ 1 10.6 10.2
ā¢ 2 9.8 9.4
ā¢ 3 12.3 11.8
ā¢ 4 9.7 9.1
ā¢ 5 8.8 8.3
ā¢
Varsha Varde 16
17. ā¢ xĀÆ1 = 10.24 xĀÆ2 = 9.76
ā¢ Using the previous section test we would have t = 0.57 resulting in
an insignificant test which is inconsistent with the data.
ā¢ Automobile Tire A Tire B d=A-B
ā¢ 1 10.6 10.2 0.4
ā¢ 2 9.8 9.4 0.4
ā¢ 3 12.3 11.8 0.5
ā¢ 4 9.7 9.1 0.6
ā¢ 5 8.8 8.3 0.5
ā¢ xĀÆ1 = 10.24 xĀÆ2 = 9.76 dĀÆ =0 .48
ā¢ Q1: Provide a summary of the data in the above table.
ā¢ Sample summary: n = 5, d ĀÆ= .48, sd = .0837
Varsha Varde 17
18. ā¢ Q2: Do the data provide sufficient evidence to indicate a
difference in average wear for the two tire types.
ā¢ Test. (parameter Āµd = Āµ1 - Āµ2)
ā¢ H0 : Āµd = 0
ā¢ Ha : Āµd ā 0
ā¢ T.S. :t =d ĀÆ- D0/sd/ān
ā¢ =.48 ā 0/.0837/ā5 = 12.8
ā¢ RR: Reject H0 if t > 2.776 or t < -2.776 ( t.025,4 = 2.776)
ā¢ Decision: Reject H0
ā¢ Conclusion: At 5% significance level there is sufficient
statistical evidence to to conclude that the average
amount of wear for type A tire is different from that for
type B tire.
ā¢ Exercise. Construct a 99% confidence interval for the
18
19. Inferences About a Population Variance
ā¢ Chi-square distribution. When a random
sample of size n is drawn from a normal
population with mean Āµ and standard deviation
Ļ, the sampling distribution of S2
depends on n.
The standardized distribution of S2
is called the
chi-square distribution and is given by
ā¢ Ļ2
=(n - 1)s2
/Ļ2
ā¢ Degrees of freedom (df): Ī½ = n - 1
ā¢ Graph: Non-symmetrical and depends on df
ā¢ Critical values: using Ļ2
tables
Varsha Varde 19
20. Inferences About a Population Variance
ā¢ Test.H0 : Ļ2
= Ļ2
0
ā¢ Ha : Ļ2
ā Ļ2
0 (two-tailed test).
ā¢ Ha : Ļ2
<Ļ2
0 (One ātailed test )
ā¢ T.S. :X2
=(n - 1)s2
/ Ļ2
0
ā¢ RR: (two-tailed test).
ā¢ Reject H0 if X2
> X2
Ī±/2 or X2
< X2
1-Ī±/2 where
ā¢ X2
is based on (n - 1) degrees of freedom.
ā¢ RR: (One-tailed test).
ā¢ Reject H0 if X2
< X2
1-Ī± where
ā¢ X2
is based on (n - 1) degrees of freedom
ā¢ Assumptions.
ā¢ 1. Normal population
ā¢ 2. Random sample
Varsha Varde 20
21. Example
ā¢ Future Technologies ltd. manufactures high
resolution telescopes .The management wants
its products to have a variation of less than 2 sd
in resolution while focusing on objects which are
beyond 500 light years. When they tested their
newly manufactured telescope for 30 times to
focus on an object 500 light years away they
found sample sd to be 1.46. State the
Hypothesis and test it at 1% level of significance.
Can the management accept to sell this
product?
Varsha Varde 21
22. Solution
ā¢ Given ,n=30 and S2
=(1.46)2
,Ļ2
0 =4We set up hypothesis
ā¢ H0: Ļ2 =4
ā¢ Ha: Ļ2< 4
ā¢ Significance Level Ī± =1%=.01
ā¢ This is a one tailed test
ā¢ T.S = X2
=(n - 1)s2
/ Ļ2
0= (30-1)(1.46)2/4=15.45
ā¢ Referring to X2
tables we find at 29 d.f the value of X2
that
leaves an area of 1-.01 =.99 in the upper tail and .01 in
the lower tail is 14.256.Since calculated value 15.45 is
greater than the cut off level 14.256 we accept null
hypothesis and conclude that sd=2. There fore
management will not allow sale of its telescope.
Varsha Varde 22
23. Comparing Two Population Variances
ā¢ F-distribution. When independent samples are drawn from two
normal populations with equal variances then S2
1/S2
2 possesses a
sampling distribution that is known as an
ā¢ F distribution. That is
ā¢ F = S2
1/S2
2
ā¢ Degrees of freedom (df): Ī½1 = n1 - 1; Ī½2 = n2 - 1
ā¢ Graph: Non-symmetrical and depends on df
ā¢ Critical values: using F tables
ā¢ Test.
ā¢ H0 : Ļ2
1 = Ļ2
2
ā¢ Ha : Ļ2
1ā Ļ2
2(two-tailed test).
ā¢ T.S. :F = S2
1/S2
2
ā¢ Where S2
1is the larger sample variance.
ā¢ Note: F = larger sample variance/smaller sample variance
ā¢ RR: Reject H0 if F > FĪ±/2 where FĪ±/2 is based on (n1 - 1) and (n2 - 1)
degrees of freedom.
ā¢ Assumptions. 23
24. Example
ā¢ Investment risk is generally measured by the volatility of
possible outcomes of the investment. The most common
method for measuring investment volatility is by
computing the variance ( or standard deviation) of
possible outcomes. Returns over the past 10 years for
first alternative and 8 years for the second alternative
produced the following data:
ā¢ Data Summary:
ā¢ Investment 1: n1 = 10, xĀÆ1 = 17.8%; s2
1 = 3.21
ā¢ Investment 2: n2 = 8, xĀÆ2 = 17.8%; s2
2 = 7.14
ā¢ Both populations are assumed to be normally
distributed.
Varsha Varde 24
25. Solution
ā¢ Q1: Do the data present suffcient evidence to indicate
that the risks for investments1 and 2 are unequal ?
ā¢ Solution.
ā¢ Test:H0 : Ļ2
1 = Ļ2
2
ā¢ Ha : Ļ2
1 ā Ļ2
2 (two-tailed test).
ā¢ T.S. :F = S2
1/S2
2 =7.14/3.21= 2.22
ā¢ .RR: Reject H0 if F > FĪ±/2 where
ā¢ FĪ±/2,n2-1,n1-1 = F.025,7,9 = 4.20
ā¢ Decision: Do not reject H0
ā¢ Conclusion: At 5% significance level there is insuffcient
statistical evidence to indicate that the risks for
investments 1 and 2 are unequal.
ā¢ Exercise. Do the upper tail test. That is
ā¢ Ha : Ļ2
1 > Ļ2
2.
25