Exercises 5

Sample size determination:

Error of estimation
𝑍 𝛼 𝑆 2 𝑍 𝛼 𝜎 2
1− 1−
2 2
𝑛=( 𝑑
) 𝑜𝑟 𝑛 = ( 𝑑
) , in case (WR).

𝑁𝜎 2 𝑑
𝑛 = (𝑁−1)𝐷2 +𝜎2 , 𝐷= 𝑧
, in case (WOR).

Coefficient of variation
1 𝑠2
𝑛=
𝐶𝑉 2 𝑦̅ 2

Relative error (Relative Differences )

𝑍 𝛼 𝑆 2 𝑍2 𝛼 𝑆 2
1− 1−
2 2
𝑛=( ) 𝑜𝑟 𝑛 = (𝑎 𝑦̅ )2
a: is error should not exceed
𝑎𝑝

[Note: We always round up; the sample size formulas always generate the minimum number of
subjects needed to ensure the specified precision.]

Example 1: (Sample size estimation for proportion in survey)

Researcher interested to know the sample size for conducting a survey for measuring the
prevalence of obesity in certain community. Previous literature gives the estimate of an
obesity at 20% in the population to be surveyed, and assuming 95% confidence interval (or
5% level of significance) and 10% Relative error. Find the sample sizes required (WR)?
The sample size can be calculated as follow as
𝑝 = 0.2 , 1 − 𝛼 = 0.95 , 𝛼 = 0.05
𝑍2 𝛼 𝑝𝑞
1−
2 1.962 ∗0.2∗(1−0.2)
𝑛= (𝑎 𝑝)2
= (0.1∗0.2)2
= 1536.64 ≈ 1537 for a simple random sampling design.

Exercises 2: (Sample size estimation for average in survey)

Consider a camera club with 1800 members, where it is required to estimate the average
number of rolls of film used during a year. Consider also the information from the past
that the average and standard deviation of the number of rolls of film have been
around 6 and 4, respectively.
Find the sample sizes required (WR)to estimate the average number of rolls of film with
following information:
 a) With an error not exceeding 1, and 5% level of significance, with the normal
approximation.
𝑍1−𝛼 𝑆 2 1.96 ∗ 4 2
2
𝑍0.975 = 1.96 , 𝑛 = ( ) = ( ) = 61.46 ≈ 62
𝑑 1

b) With C.V. should not exceed 8%.

𝑠
With the information on the mean and standard deviation, 𝑐𝑦 = 𝑐. 𝑣 = 𝑦̅
1 𝑠2 𝑐. 𝑣 2 1 4 2
𝑛= = = ( ) = 69.44 ≈ 70
𝐶𝑉 2 𝑦̅ 2 𝐶𝑉 2 0.082 6

c) If the relative error should not exceed (0.1), except for α=0.05.
2 2
𝑍1− 𝛼 𝑆
2 (1.96 ∗ 4)2
𝑛= = = 170.73 ≈ 171
(𝑎 𝑦̅ )2 (0.1 ∗ 6)2

Exercises 3: (Sample size estimation for average in survey)

The metropolitan area and the suburbs together in a region consist of 5, 10, and 10
thousand families with one, two, and three or more children. For these three types of
families, preliminary estimates of the averages and standard deviations of the number of
hours of television watching in a week are (10, 15, 20) and (6, 10, 15), respectively.
Find the sample sizes required (WR) to estimate the above average for each group if the
error of estimation should not exceed 2 hours in each case, except for α=0.05.
𝑍1−𝛼 𝑆 2
2
𝑍0.975 = 1.96 , 𝑛 = ( )
𝑑

1.96 ∗ 6 2
𝑛1 = ( ) = 34.56 ≈ 35
2
1.96 ∗ 10 2
𝑛2 = ( ) = 96.04 ≈ 97
2
1.96 ∗ 15 2
𝑛3 = ( ) = 216.09 ≈ 217
2
We note the sample size are differences, because the standard errors are different from each
group.
Exercises 4. (Sample size estimation for proportion in survey). (H.W)
One university has 1000 students in each of the four classes. The percentages of the
Freshman–Sophomore, Junior, and Senior classes expressing interest in professional
training after graduation were guessed to be 20, 50, and 80%, respectively.

(a) For each of these three groups, find the sample sizes required to estimate the
percentage if the estimate should not differ from the actual value by not more than
20% of the actual value except for α=0.05, and present the reason for the differences
in the sample sizes.

(b) Find the sample sizes needed for each of the three groups for estimating the above
percentage if the error of estimation should not exceed 10% except for α=0.05 and
present the reason for the differences in the sample sizes.

Exercises 5 : A survey is to be conducted to estimate the average monthly

income of a locality with 5000 households. It is known that the squared
variability in the income is 250000 Rayals. How many households should we
select (WOR) so that the marginal error of estimated income is no more than
200 Rayals? (use α=0.05)

Solution: we have N= 5000, σ2 =250000 and d=200

𝑁𝜎 2 𝑑
𝑛= , 𝐷=
(𝑁 − 1)𝐷2 + 𝜎 2 𝑧
200 200 5000∗250000
𝐷=𝑍 = 1.96 = 102.0408 , 𝑛 = 4999∗102.04082+250000 = 23.9 ≈ 24
0.975

So, a sample of 24 householder is required to conduct the survey.

Exercises 6: It is known that the proportion of smokers in a society of 4000
individuals is 0.2. How many individuals we need to select (WOR) if we want
to estimate the proportion of smokers with error in estimated “marginal error”
equal to 4%. (use α=0.05)
Solution: we have N= 4000, d=0.04 , P =0.2 so Q=1-0.2=0.8
The sample size for estimation of population mean is given as
𝑁𝜎 2 𝑑
𝑛= 2 2
, 𝐷=
(𝑁 − 1)𝐷 + 𝜎 𝑧
Further, the sample size for estimation of population proportion is given as
𝑁𝑃𝑄
𝑛=
(𝑁 − 1)𝐷2 + 𝑃𝑄
2
2
𝑑 0.04 2
𝐷 =( ) =( ) = 0.0004
𝑍0.975 1.96
𝑁𝑃𝑄 4000 ∗ 0.2 ∗ 0.8
𝑛= 2
= = 363.72 ≈ 364
(𝑁 − 1)𝐷 + 𝑃𝑄 (3999 ∗ 0.0004) + (0.2 ∗ 0.8)
Hence, we need a sample of 364 individuals to estimate the proportion of smokers.
 Exercises 6
#Cost -sample size
Example 1: student was asked to take up the problem about estimate the average time per
week devoted to study in University library by the students of this university. He was
provided with $150, including overhead cost of $ 24. The cost of contacting the students,
and collecting information is $ 3 per student. How many students would he select in the
sample, for collecting the desired information.
C(n) = cost of taking n samples =150
co = fixed cost= 24
c1 = cost for each sample interview
𝐶(𝑛) = 𝑐𝑜 + 𝑛 ∗ 𝑐1
𝐶(𝑛) − 𝑐𝑜 150 − 24
𝑛= = = 42
𝑐1 3
#Power in sample size
Power (1 − 𝜷) 𝒁𝟏−𝜷 Alpha 𝒁𝟏−𝜶
𝟐
0.70 0.524 0.2 1.281
0.75 0.674 0.15 1.44
0.80 0.841 0.10 1.645
0.85 1.036 0.05 1.960
0.90 1.282 0.01 2.576
0.95 1.645 0.001 3.29
0.99 2.326

NOTE:
» The higher power will require a larger sample size.
» Sample Size for One Sample, Continuous Outcome 𝐻0 : 𝜇 = 𝜇1 𝑣𝑠 𝐻1 : 𝜇 ≠ 𝜇1
2
(𝑍1−𝛼 + 𝑍1−𝛽 ) 𝑆 2
2
𝑛= , 𝑑 = 𝜇1 − 𝜇0
(𝑑)2

»Sample Sizes for Two Independent Samples, Continuous Outcome 𝐻0 : 𝜇1 = 𝜇2 𝑎𝑛𝑑 𝐻1 : 𝜇1 ≠ 𝜇2

under the assumption of common variance (i.e equal variability in the two populations)
2
2 (𝑍1−𝛼 + 𝑍1−𝛽 ) 𝑆 2
2
𝑛= , 𝑑 = 𝜇1 − 𝜇2
(𝑑)2

»Sample Sizes for Two Independent Samples, Dichotomous Outcomes 𝐻0 : 𝑝1 = 𝜇2 𝑣𝑠 𝐻1 : 𝜇1 ≠ 𝜇2

2
2 (𝑍1−𝛼 + 𝑍1−𝛽 ) 𝑝ҧ𝑞̅
2
𝑛=
(𝑝1 − 𝑝2 )2
Where 𝑝ҧ is mean of the proportions in the two comparison groups, assuming that the groups will be of
approximately equal size.
Example 2: An investigator wishes to compare two treatments for nausea, one being
placebo and the other being a new experimental drug. The absolute risk of nausea on
placebo is predicted to be 50% and it is thought that the new treatment would be worth
using if it reduced the absolute risk of nausea to 30%, meaning that the treatment effect
would have an absolute risk reduction of 20%. The trial will have 90% power and a two-
sided significance level of 5%.
How many students would he select in the sample? (Two-sided test)
2
(𝑍1−𝛼 + 𝑍1−𝛽 ) 2𝑝ҧ 𝑞̅ 𝑝1 + 𝑝2
2
𝑛= ; 𝑝ҧ =
(𝑝1 − 𝑝2 )2 2
𝑝1 + 𝑝2 0.5 + 0.3
𝑝ҧ = = = 0.4 , 𝑞̅ = 1 − 0.4 = 0.6
2 2
𝑍1−𝛼 = 𝑍0.975 = 1.96 ; 𝑍1−𝛽 = 𝑍0.90 = 1.282
2

(1.96 + 1.282)2 2 ∗ 0.4 ∗ 0.6

𝑛= = 126.126 ≈ 127 𝑝𝑒𝑟 𝑔𝑟𝑜𝑢𝑝
(0.2)2
𝑛1 = 127, 𝑛2 = 127

Example 3: It is believed that the proportion of patients who develop complications after
undergoing one type of surgery is 5% while the proportion of patients who develop
complications after a second type of surgery is 15%. How large should the sample be in
each of the 2 groups of patients if an investigator wishes to detect, with a power of 90%,
and 5% level of significance? (Two-sided test)
2
(𝑍1−𝛼 + 𝑍1−𝛽 ) 2𝑝ҧ 𝑞̅ 𝑝1 + 𝑝2
2
𝑛= ; 𝑝ҧ =
(𝑝1 − 𝑝2 )2 2
𝑝1 + 𝑝2 0.05 + 0.15
𝑝ҧ = = = 0.10 , 𝑞̅ = 1 − 0.1 = 0.90
2 2
𝑍1−𝛼 = 𝑍0.975 = 1.96 ; 𝑍1−𝛽 = 𝑍0.90 = 1.282
2

(1.282 + 1.96)2 ∗ 2 ∗ (0.10) ∗ (0.90)

𝑛= = 189.19 = 190;
(0.05 − 0.15)2
Example 4: (H.W)
An investigator wishes to compare two treatments for patients who’ve suffered from a
heart attack. It is known that 20% of people in placebo group die within one year, while
5% of people in treatment group die within one year.
How large should the sample be in each of the 2 groups, with a power of 80%, and 5%
level of significance? (Two-sided test)

Example 5: (Sample Sizes for One Samples)

In a study for estimating the weight of population and wants the error of estimation to be
less than 2 kg of true mean (that is expected difference of weight to be 2 kg), the sample
standard deviation was 5 kg. Find the sample size required for this study, with statistical
power of 90% at 5% level significance.
the sample size estimated as
2
(𝑍1−𝛼 + 𝑍1−𝛽 ) 𝑆 2
2
𝑛=
(𝑑)2
𝑍1−𝛼 = 𝑍0.975 = 1.96 ; 𝑍1−𝛽 = 𝑍0.90 = 1.282
2

(1.96 + 1.282)2 52
𝑛= = 65.69 ≈ 66
22
Example 6: (Sample Sizes for Two Independent Samples, Continuous Outcome)
An investigator is planning a clinical trial to evaluate the efficacy of a new drug designed
to reduce blood pressure. The plan is to enroll participants and to randomly assign them to
receive either the new drug or a placebo. If the new drug shows a 5 unit reduction in mean
blood pressure, this would represent a clinically meaningful reduction.
How many patients should be enrolled in the trial to ensure that the power of the test is
80% to detect this difference? A two-sided test will be used with a 5% level of significance.
In addition to, the standard deviation of blood pressure from previous Heart Study was 19
unit.
2
𝟐 (𝑍1−𝛼 + 𝑍1−𝛽 ) 𝑆 2
2
𝑛=
(𝑑)2
𝑍1−𝛼 = 𝑍0.975 = 1.96 ; 𝑍1−𝛽 = 𝑍0.80 = 0.84
2

(1.96 + 0.84)2 ∗ 2 ∗ (19)2

𝑛= = 226.4 ≈ 227
(5)2

#Sampling weighting
Determine weighting for the following sampling designs:
𝑁 20000
1) If N=20000 , n= 400 . >>> 𝑤 = = = 50
𝑛 400

𝑁 1559
2) If N=1559 , n= 60 . >>> 𝑤 = = = 25.983
𝑛 60

𝑁 335
3) If N=335 , n= 230 . >>> 𝑤 = = 230 = 1.456
𝑛