One – Population Case

Table 1. Tests Concerning the Population Mean

Ho:  = o
(where o is a specified hypothesized value of the population mean)
Test Statistic Ha Region of Rejection

Case 1:  known  < o zc < -z

 > o zc > z
  o zc<-z/2 & zc >z/2

Case 2:  unknown  < o tc < -t(, v)

 > o tc > t(, v)
  o tc <-t (/2, v)
& tc >t (/2, v)
v=n-1

◼ The above tests are exact -level tests for samples from a normal distribution. However, they provide good
approximate -level test when the distribution is not normal provided that the sample size is n>30.
◼ Case 3: If 2 is unknown and n>30, use the z-test but replace  by s, that is,

x − o
Zc =
s/ n
Note: tabulated z-values for the common choices of 
 0.01 0.05 0.10

z 2.33 1.645 1.28

z/2 2.576 1.96 1.645

One-Tail Test

Question 1: Does an average box of cereal contain more than 368 grams of cereal? A random sample of 25 boxes
showed 372.5 grams. The company has specified 𝜎 to be 15 grams. Test at the a = 0.05 level.
H0: 𝜇 = 368 𝑔
Ha: 𝜇 > 368 𝑔
where = true mean content of cereals

Population standard deviation, 𝜎, is know so use 𝑧.

𝑋̅−𝜇 372.5−368
Test Statistic: 𝑧𝑠𝑡𝑎𝑡 = = = 1.5
𝜎/√𝑛 15/√25
𝑧𝑐𝑟𝑖𝑡 = 1.645
Decision: Since 𝑧𝑠𝑡𝑎𝑡 = 1.5 < 1.645, we fail to reject 𝐻0 at 𝛼 = 0.05.
Conclusion: The sample does not provide sufficient evidence to support that the true mean is more than 368.

One-Tail Test
Question 2: Does an average box of cereal contain more than 368 grams of cereal? A random sample of 25 boxes
showed 372.5 grams. The company has specified 𝑠 to be 15 grams. Test at the a = 0.05 level.
H0: 𝜇 = 368 𝑔
Ha: 𝜇 > 368 𝑔
where = true mean content of cereals

Population standard deviation, 𝜎, is unknown and 𝑛 ≤ 30, use 𝑡.

𝑋̅−𝜇 372.5−368
Test Statistic: 𝑡𝑠𝑡𝑎𝑡 = 𝑠/ = = 1.5
√𝑛 15/√25
𝑡𝑐𝑟𝑖𝑡 = 1.71
Decision: Since 𝑧𝑠𝑡𝑎𝑡 = 1.5 < 1.71 , we fail to reject 𝐻0 at 𝛼 = 0.05.
Conclusion: The sample does not provide sufficient evidence to support that the true mean is more than 368.

Two-Tail Test
Question 2: Does an average box of cereal is not equal to 368 grams of cereal? A random sample of 40 boxes showed
372.5 grams. The company has specified 𝑠 to be 15 grams. Test at the a = 0.05 level.
H0: 𝜇 = 368 𝑔
Ha: 𝜇 ≠ 368 𝑔
where = true mean content of cereals

Population standard deviation, 𝜎, is unknown but 𝑛 > 30, use 𝑧.

𝑋̅−𝜇 372.5−368
Test Statistic: 𝑧𝑠𝑡𝑎𝑡 = 𝑠/ = = 1.897
√𝑛 15/√40
𝑧𝑐𝑟𝑖𝑡 = 1.645
Decision: Since 𝑧𝑠𝑡𝑎𝑡 = 1.897 > 1.645, we reject 𝐻0 at 𝛼 = 0.05.
Conclusion: The sample provides sufficient evidence to support that the true mean is not equal to 368g.

Two – Population Case

Ho: X - Y = 0 The means of the 2 populations

are the same.
Possible Alternatives (Ha):
X - Y < 0 The mean of Population 1 is smaller
than the mean of Population 2.
X - Y > 0 The mean of Population 1 is larger
than the mean of Population 2.
X - Y  0 The two population means are
different.

Two Independent Samples Two Related Samples or Paired Observations

◼ the selection of the sample in ◼ achieved by either using the same subject in the two
Population 1 is independent of the samples or pairing of subjects with respect to some
selection of the sample in Population 2 extraneous variable that may affect or influence the
outcome
◼ used to overcome the difficulty imposed by extraneous
differences between the two populations

Testing the Difference Between Two Population Means Based On Two Independent Samples
Ho: X - Y = 0
Test statistic Ha Region of
Rejection

Case 1: 𝑋 & 𝑌 known 𝑋 − 𝑌 < 0 𝑧𝑠𝑡𝑎𝑡 < −𝑧

𝑧𝑠𝑡𝑎𝑡 =
𝑋̅−𝑌̅ 𝑋 − 𝑌 > 0 𝑧𝑠𝑡𝑎𝑡 > 𝑧
2 2
√𝜎𝑥 +𝜎𝑌 𝑋 − 𝑌  0 𝑧𝑠𝑡𝑎𝑡 < −𝑧𝛼 & 𝑧𝑠𝑡𝑎𝑡 > 𝑧𝛼
𝑛 𝑚 2 2

Case 2: 𝑋 & 𝑌 unknown but equal 𝑋 − 𝑌 < 0 𝑡𝑠𝑡𝑎𝑡 < −𝑡(,𝑣)

𝑡𝑠𝑡𝑎𝑡 =
𝑋̅−𝑌̅ 𝑋 − 𝑌 > 0 𝑡𝑠𝑡𝑎𝑡 > 𝑡(,𝑣)
1 1
√𝑠𝑝2 ( + )
𝑛 𝑚 𝑋 − 𝑌  0 𝑡𝑠𝑡𝑎𝑡 < − 𝑡(,𝑣)
2
Where & 𝑡𝑠𝑡𝑎𝑡 > 𝑡(,𝑣)
(𝑛 − 1)𝑠𝑋2 + (𝑚 − 1)𝑠𝑌2 2
𝑠𝑝2 = 𝑣 = 𝑛 + 𝑚– 2
𝑛+𝑚−2
 Case 3: 𝑋 & 𝑌 unknown & unequal 𝑋 − 𝑌 < 0 𝑡𝑠𝑡𝑎𝑡 < −𝑡(,𝑣)
𝑋̅ − 𝑌̅ 𝑋 − 𝑌 > 0 𝑡𝑠𝑡𝑎𝑡 > 𝑡(,𝑣)
𝑡𝑠𝑡𝑎𝑡 =
2 2 𝑋 − 𝑌  0 𝑡𝑠𝑡𝑎𝑡 < − 𝑡(,𝑣)
√𝑠𝑋 + 𝑠𝑌 2
𝑛 𝑚 & 𝑡𝑠𝑡𝑎𝑡 > 𝑡(,𝑣)
2
𝑤ℎ𝑒𝑟𝑒
2
𝑠2 𝑠2
( 𝑛𝑋 + 𝑚
𝑌)

𝑣 = 2 2 2
(𝑠2
𝑋 ) +(𝑠𝑌 )
𝑛−1 𝑚−1

◼ These tests are exact -level tests for independent samples selected from normal populations. However, they
provide good approximate -level tests when the distributions are not Normal provided both samples are
greater than 30.
◼ Case 4: If X and Y are unknown and both sample sizes are greater than 30, use the z-test statistic but replacing
X2 and Y2 by sX2 and sY2, respectively.
◼ If there is no information at all about the population variances, the test based on the t-test statistic will still
provide a good approximate -level test so long as the sample sizes are the same and both populations are
Normal. This is the reason why the researcher should plan the experiment so that n=m.

Example: Testing the Difference Between Two Population Means (Paired Samples)
Subjects are tested for reactions times with their left and right hands. (Only right-handed subjects were used.) The
results (in thousandths of a second) are given in the accompanying table. Use a 0.05 significance level to test the claim
that there is a difference between the mean of the right- and left-hand reaction times. If an engineer is designing a
fighter-jet cockpit and must locate the ejection-seat activator to be accessible to either the right or the left hand, does it
make a difference which hand she chooses?
Subject A B C D E F G H I J K L M N
Right 191 97 116 165 116 129 171 155 112 102 188 158 121 133
Left 224 171 191 207 196 165 177 165 140 188 155 219 177 174

Solution:
𝐻0 : 𝜇1 = 𝜇2
𝐻𝐴 : 𝜇1 ≠ 𝜇2 (original claim)

Where 𝜇1 is the mean reaction time of right-hand persons

𝜇1 is the mean reaction time of left-hand persons

Because we are testing a claim about the means of paired dependent data, we use the Student t distribution.

t-Test: Paired Two Sample for Means

Right Left
Mean 139.5714286 182.0714286
Variance 984.4175824 567.456044
Observations 14 14
Pearson Correlation 0.302012137
Hypothesized Mean Difference 0
df 13
t Stat -4.793724854
P(T<=t) one-tail 0.000175383
t Critical one-tail 1.770933396
P(T<=t) two-tail 0.000350766
t Critical two-tail 2.160368656

Conclusion/Interpretation:
Because the test statistic does fall in the critical region, we reject the null hypothesis. There is sufficient evidence to
support the claim of a difference between the right- and left-hand reaction times.
 Because there does appear to be such a difference, an engineer designing a fighter-jet cockpit should locate the
ejection-seat activator so that it is readily accessible to the faster hand, which appears to be the right hand with
seemingly lower reaction times. (We could require special training for left-handed pilots if a similar test of left-handed
pilots show that their dominant hand is faster.)

See Excel file for other examples.