Test Of Hypothesis Involving Means, Variances, and Proportion (Miller, 2004:

398 – 423)

Traditionally, it has been the costume to outline tests of hypotheses by means of the
following steps:

1. Formulate Ho and H1, and specify α.

2. Using the sampling distribution of an appropriate test statistic, determine a critical
region of size α.
3. Determine the value of the test statistic from the sample data.
4. Check whether the value of the test statictic falls into the critical region and,
accordingly, reject the null hypotheses, or accept it, or reserve judgement.
(p. 400)

Untuk menguji hipotesis, lakukan langkang-langkah berikut:

1. Rumuskan Ho dan H1, dan tentukan α.

2. Dengan menggunakan distribusi sampling dari uji statistik yang sesuai, tentukan
daerah kritis dari ukuran α.
3. Tentukan nilai uji statistik dari data sampel.
4. Periksa apakah nilai dari uji statistik jatuh ke dalam daerah kritis dan berdasarkan itu,
tolak atau terima hipotesis nihil, atau menundanya.
Contoh pengujian 1:

Suppose that 100 tires made by a certain manufacturer lasted on the average 21,819 miles with
a standar deviation of 1,295 miles. Test the null hipothesis µ = 22,000 miles against the
alternative hypothesis µ < 22,000 miles at the 0.05 level of significance. (Miller, 2004: 404)

Solution

1. Ho: µ = 22,000
H1: µ < 22,000
α = 0.05
2. Reject the null hipothesis if 𝑧 ≤ −1.645, di mana
𝑥̅ − 𝜇0
𝑧= 𝜎
⁄ 𝑛
√
3. Substituting 𝑥 = 21,819, 𝜇0 = 22,000, 𝑠 = 1,295 𝑓𝑜𝑟 𝜎, 𝑎𝑛𝑑 𝑛 = 100, we get
21,819 − 22,000
𝑧= = −1.40
1,295
⁄
√100
4. Since z = -1.40 is greater than -1.645, the null hypothesis cannot be rejected; there is
no real evidance that the tires are not as good as assumed under the null hypothesis.

Contoh pengujian 2:

The specifications for certain kind of ribbon call for a mean breaking strength of 185 pounds.
If five pieces randomly selected from different rolls have breaking strenghts of 171.6, 191.8,
178.3, 184.9, and 189.1 pounds, test the null hypothesis µ = 185 pounds against the alternative
hypothesis µ < 185 pounds at the 0.05 level of significance.

Solution

1. Ho: µ = 185
H1: µ < 185
α = 0.05
2. Reject the null hypothesis if t ≤ −2.132, where t is determined by means of the
formula given above and 2.132 is the value of t0.05,4.
3. First we calculate the mean and the standard deviation, getting 𝑥̅ = 183.1 and s = 8.2.
Then, substituting these values together with μ0 = 185 and n = 5 into the formula for t,
we get

𝑥̅ − 𝜇0 183.1− 185
𝑡= 𝑠 = 8.2 = −0.506
⁄ 𝑛 ⁄
√ √5

4. Since t = −0.49 is greater than −2.132, the null hypothesis cannot be rejected. If we
went beyond this and concluded that the rolls of ribbon from which the sample was
selected meet specifications, we would, of course, be exposed to the unknown risk of
committing a type II error.
Contoh pengujian 3:

The following are the average weekly losses of work-hours due to accidents in 10
industrial plants before and after a certain safety program was put into operation:

45 and 36, 73 and 60, 46 and 44, 124 and 119, 33 and 35, 57 and 51, 83 and 77, 34 and
29, 26 and 24, and 17 and 11

Solution

1. Ho: 𝜇𝐷 = do
H1: 𝜇𝐷 > do
α = 0.05
2. Reject the null hipothesis if t > t ;
3. First we calculate the mean and the standard deviation, getting 𝑑̅ = 5.2 and Sd =. Then,
substituting these values together with μ0 = 185 and n = 10 into the formula for t, we get
d  d0
t =
sd
n

since t = ……., the null hypothesis is rejected; there is real evidance that the breaking strength
of the ribbon is as good as assumed under the null hypothesis.
Contoh pengujian 4:
Suppose that it is known from experience that the standard deviation of the weight of 8-ounce
packages of cookies made by a certain bakery is 0.16 ounce. To check whether its production
is under control on a given day, that is, to check whether the true average weight of the
packages is 8 ounces, employees select a random sample of 25 packages and find that their
mean weight is 𝑥̅ = 8.091 ounces. Since the bakery stands to lose money when μ>8 and the
customer loses out when μ<8, test the null hypothesis μ = 8 against the alternative hypothesis
𝜇 ≠ 8 at the 0.01 level of significance.
Solution
1. H0: μ = 8
H1: 𝜇 ≠8
α = 0.01
2. Reject the null hypothesis if z ≤ −2.575 or z ≥ 2.575, where
𝑥̅ −𝜇0
𝑧=
𝜎/√𝑛

3. Substituting x = 8.091, μ0 = 8, σ = 0.16, and n = 25, we get

8.091 − 8
𝑧=
0.16/√25

= 2.84

4. Since z = 2.84 exceeds 2.575, the null hypothesis must be rejected and suitable
adjustments should be made in the production process.
EXAMPLE 4

An experiment is performed to determine whether the average nicotine content of one kind of
cigarette exceeds that of another kind by 0.20 milligram. If n1 = 50 cigarettes of the first kind
had an average nicotine content of x1 = 2.61 milligrams with a standard deviation of s1 = 0.12
milligram, whereas n2 = 40 cigarettes of the other kind had an average nicotine content of x2
= 2.38 milligrams with a standard deviation of s2 = 0.14 milligram, test the null hypothesis μ1
−μ2 = 0.20 against the alternative hypothesis μ1 −μ2 ≠ 0.20 at the 0.05 level of significance.
Base the decision on the P-value corresponding to the value of the appropriate test statistic.

Solution

1. H0: μ1 −μ2 = 0.20

H1: μ1 −μ2 ≠ 0.20
α = 0.05

2. Use the test statistic Z, where

𝑥̅1 − 𝑥̅ 2 − 𝛿
𝑧=
𝜎2 𝜎2
√ 1 + 2
𝑛1 𝑛2

3. Substituting 𝑥̅1 = 2.61, 𝑥̅ 2 = 2.38, δ = 0.20, s1 = 0.12 for σ1, s2 = 0.14 for σ2, n1 = 50, and n2
= 40 into this formula, we get

2.61 − 2.38 − 0.2

𝑧= = 1.08
2 2
√(0.12) + (0.14)
50 40

The corresponding P-value is 2(0.5000−0.3599) = 0.2802, where 0.3599 is the entry in

Table III of “Statistical Tables” for z = 1.08.

5. Since 0.2802 exceeds 0.05, the null hypothesis cannot be rejected; we say that the
difference between 2.61−2.38 = 0.23 and 0.20 is not significant. This means that the
difference may well be attributed to chance. When n1 and n2 are small and σ1 and σ2 are
unknown, the test we have been discussing cannot be used. However, for independent
random samples from two normal populations having the same unknown variance σ2, the
likelihood ratio technique yields a test based on

𝑥̅1 − 𝑥̅ 2 − 𝛿
𝑡=
1 1
𝑠𝑝 √𝑛 + 𝑛
1 2

(𝑛1− 1)𝑠12+(𝑛2− 1)𝑠22

where 𝑠𝑝2 = 𝑛1 +𝑛2 −2

Under the given assumptions and the null hypothesis μ1 −μ2 = δ, this expression for t is a
value of a random variable having the t distribution with n1 + n2 −2 degrees of freedom. Thus,
the appropriate critical regions of size α for testing the null hypothesis μ1 −μ2 = δ against the
alternatives μ1 −μ2 ≠ δ, μ1 −μ2 >δ, or μ1 −μ2 < δ under the given assumptions are, respectively,
|t| ≥ tα/2, 𝑛1 + 𝑛2 − 2, t ≥ tα, 𝑛1 + 𝑛2 − 2, and t ≤ −tα, 𝑛1 + 𝑛2 − 2. To illustrate this two-
sample t test, consider the following problem.

EXAMPLE 5

In the comparison of two kinds of paint, a consumer testing service finds that four 1-gallon
cans of one brand cover on the average 546 square feet with a standard deviation of 31 square
feet, whereas four 1-gallon cans of another brand cover on the average 492 square feet with a
standard deviation of 26 square feet. Assuming that the two populations sampled are normal
and have equal variances, test the null hypothesis μ1 −μ2 = 0 against the alternative hypothesis
μ1 −μ2 > 0 at the 0.05 level of significance.

Solution

1. H0: μ1 −μ2 = 0
H1: μ1 −μ2 > 0
α = 0.05

2. Reject the null hypothesis if t ≥ 1.943, where t is calculated according to the formula given
on the previous page and 1.943 is the value of t0.05,6.

3. First calculating sp, we get

3(31)2 + 3(26)2
𝑠𝑝 = √
4+4−2

= 28.609

and then substituting its value together with 𝑥̅1 = 546, 𝑥̅ 2 = 492, δ = 0, and n1 = n2 = 4 into the
formula for t, we obtain

546 − 492 − 0
𝑡=
1 1
28.609√4 + 4

= 2.67

4. Since t = 2.67 exceeds 1.943, the null hypothesis must be rejected; we conclude that on the
average the first kind of paint covers a greater area than the second. Note that n1 = n2 in
this example, so the formula for 𝑠𝑝2 becomes

1
𝑠𝑝2 = (𝑠12 + 𝑠22 )
2

Use of this formula would have simplified the calculations in this special case. In Exercise 41
the reader will be asked to use suitable computer software to show that the P-value would
have been 0.0185 in this example, and the conclusion would, of course, have been the same.
If the assumption of equal variances is untenable in a problem of this kind, there are several
possibilities. A relatively simple one consists of randomly pairing the values obtained in the
two samples and then looking upon their differences as a random sample of size n1 or n2,
whichever is smaller, from a normal population that, under the null hypothesis, has the mean
μ = δ. Then we test this null hypothesis against the appropriate alternative by means of the
methods of Section 2. This is a good reason for having n1 = n2, but there exist alternative
techniques for handling the case where n1 Z n2 (one of these, the Smith–Satterthwaite test, is
mentioned among the references at the end of the chapter).
So far we have limited our discussion to random samples that are independent, and the
methods we have introduced in this section cannot be used, for example, to decide on the
basis of weights “before and after” whether a certain diet is really effective or whether an
observed difference between the average I.Q.’s of husbands and their wives is really
significant. In both of these examples the samples are not independent because the data are
actually paired. A common way of handling this kind of problem is to proceed as in the
preceding paragraph, that is, to work with the differences between the paired measurements
or observations. If n is large, we can then use the test described in Section 2 to test the null
hypothesis μ1 −μ2 = δ against the appropriate alternative, and if n is small, we can use the t
test described also in Section 2 provided the differences can be looked upon as a random
sample from a normal population.

EXAMPLE 7

In comparing the variability of the tensile strength of two kinds of structural steel, an
experiment yielded the following results: 𝑛2 =13, 𝑠12 = 19.2, 𝑛2 = 16, and 𝑠22 = 3.5, where the
units of measurement are 1,000 pounds per square inch. Assuming that the measurements
constitute independent random samples from two normal populations, test the null hypothesis
𝜎12 =𝜎22 against the alternative hypothesis 𝜎12 ≠ 𝜎22 at the 0.02 level of signififance.

Solution

1. H0: 𝜎12 =𝜎22

H1: 𝜎12 ≠ 𝜎22
α = 0.02

1. Since s21
G s22
, reject the null hypothesis if
s21
s22
G 3.67, where 3.67 is the value of
4. f0.01,12,15.

5. First calculating sp, we get

3(31)2 + 3(26)2
𝑠𝑝 = √
4+4−2

= 28.609
and then substituting its value together with 𝑥̅1 = 546, 𝑥̅ 2 = 492, δ = 0, and n1 = n2 = 4 into the
formula for t, we obtain

546 − 492 − 0
𝑡=
1 1
28.609√4 + 4

= 2.67

4. Since t = 2.67 exceeds 1.943, the null hypothesis must be rejected; we conclude that on the
average the first kind of paint covers a greater area than the second. Note that n1 = n2 in
this example, so the formula for 𝑠𝑝2 becomes

1
𝑠𝑝2 = (𝑠12 + 𝑠22 )
2
13.19. Based on certain data, a null hypothesis is rejected at the 0.05 level of significance.
Would it also be rejected at the

(a) 0.01 level of significance;

(b) 0.10 level of significance?

13.20. In the test of certain …….. hypothesis is rejected at the 0.05 level of
significance. Would it also be rejected at the
(a) 0.01 level of significance;
(b) 0.10 level of significance?
EXAMPLE 10

Determine, on the basis of the sample data shown in the following table, whether the true proportion of
shoppers favoring detergent A over detergent B is the same in all three cities:

Number favoring Number favoring

ditergent A ditergent B
Los Angeles 232 168 400
San Diego 260 240 500
Fresno 197 203 400

Use the 0.05 level of significance.

(Miller & Miller, 2008: 376)

Solution
1. H0: θ1 = θ2 = θ3
H1: θ1, θ2, and θ3 are not all equal.
α = 0.05
2. Reject the null hypothesis if χ2 G 5.991, where
χ2 =
_3
i=1
_2
j=1
(fij −eij)2
eij
and 5.991 is the value of χ2
0.05,2.
3. Since the pooled estimate of θ is
θˆ = 232+260+197
400+500+400
= 689
1,300
= 0.53
the expected cell frequencies are
e11 = 400(0.53) = 212 and e12 = 400(0.47) = 188
e21 = 500(0.53) = 265 and e22 = 500(0.47) = 235
e31 = 400(0.53) = 212 and e32 = 400(0.47) = 188
and substitution into the formula for χ2 given previously yields
χ2 = (232−212)2
212
+ (260−265)2
265
+ (197−212)2
212
+ (168−188)2
188
+ (240−235)2
235
+ (203−188)2
188
= 6.48
4. Since χ2 = 6.48 exceeds 5.991, the null hypothesis must be rejected; in other
words, the true proportions of shoppers favoring detergent A over detergent B
in the three cities are not the same.