Sample Size Estimation using Yamane and Cochran and Krejcie

and Morgan and Green Formulas and Cohen Statistical Power

Analysis by G*Power and Comparisons

Chanuan Uakarn1
Kajohnsak Chaokromthong2
Nittaya Sintao3
Kasem Bundit University1
Rajamangala University of Technology Suvarnabhumi2
Stamford International University3
E-mail: chanuan.uak@kbu.ac.th1
E-mail: kajohnsak7@gmail.com2
E-mail: nitsin07@gmail.com3

Received: December 1, 2021; Revised: December 14, 2021; Accepted: December 20, 2021

ABSTRACT
Sample Size Determination of Quantitative Research; population size as research fields
for the researchers to collect data for analyzing Descriptive Statistics and Inferential Statistics.
Therefore sample size refers to a representative of the population. Research Purposes are: to
study sample size determination of academic officer for references to determine the sample
size of the research appropriate for population size, and to calculate sample size by G*Power.
Derivation of the sample size of this research is Documentary Research by calculating from
various formulas of the academic officer such as Taro Yamane, Krejcie & Morgan, Cochran,
Green, S. B., and Cohen J. which including calculation formulas, table, comparison and
computer program of G*Power - depending on variable types, variable measurement scales,
Type I error (Alpha, 𝛼=0.05), Type II error (Beta, 𝛽), Power of Test: 1 − 𝛽 =0.80 and Effect
Size=0.30.

KEYWORDS: Sample size, Population, Significance, Size of a test, Effect size, Power of
test.

Introduction Classified by scope of population into 2

In Quantitative Research; it is tough types such as 2.1 Finite population is every
for researchers to access a large population unit of study can specify scope or
size; therefore, the researchers need to completely count all the numbers 2.2
reduce the population size into correct, Infinite Population is every unit of study but
adequate and appropriate sample size for this cannot specify scope or completely
collecting data from research fields by count all the numbers and 3) Classified by
processing reference number analyzed from population characteristics into 2 types
sample called Statistics back to Parameters (Sedlack & Stanley, 1992) such as 3.1
called population. Population has classified Homogeneity is population in every unit has
into 3 types as follows 1) Classified by similar structural features 3.2 Heterogeneity
group into 2 groups such as 1.1 General is population in each unit has different
population 1.2 Hypothetical population 2) structural features.

Page 76 : APHEIT INTERNATIONAL JOURNAL

 Sample is a representative as part of - Tens Thousands of populations;
the population that the researchers are use 5-10% of sample size
interested in. A good sample is the sample - Hundreds Thousands of
with complete important characteristics that population; use 1-5% of sample
are similar to the population, and is also a size
good representative of the population. 2. Sample size determination by
calculation formulas.
Good sample characteristics
3. Sample size determination by
A Good and reliable sample as a
using tables.
representative of the population in every
4. Sample size determination by
research shall have characteristics as follows
calculating computer programs.
1. Sample must have the same
characteristics as population. The more of
similarity, numbers of research will not be Purposes
required. On the other hand, the fewer of 1. To study sample size
similarity, many more numbers of research determination of academic officer for
will not be required. references.
2. Sample should derive by 2. To determine sample size of the
probability sampling in accordance with research appropriate for population size.
preliminary agreement or conditions in 3. To calculate sample size by
statistic that is used for data analysis, G*Power.
especially Inferential Statistics that is used Benefit of Research
for referencing research results to 1. Researchers can determine the
population; there will be most preliminary correct sample size.
agreement or conditions that sample should 2. Guidelines of sample size
derive by probability sampling. selection or determination of different
Theory methods.
Sample Size Determination; the 3. Facilities for researchers by using
researchers should consider various factors computer tools.
as follows; Research Process
1) Expense, time, labor and data A researchers conduct a research by
collection tools.
Documentary Research.
2) Population size. 1. Population and sample
3) Similarity; the more of similar determination; researchers conduct a
population, the usage of samples size will be research by documentary research in order
small. If the population is very different,
to search for the truth of the phenomenon
there will be plenty of variances, therefore, of sample size determination that
the usage of sample size will be large. appropriates for the real representatives of
4) Accuracy population size in Quantitative Research
5) Sampling error (Chanuan, 2020).
6) Reliability 2. Creation of data collection tools
Sample size determination: which types of source document;
researchers collect data from both national
1. Sample size determination by
and international researches as published
using criteria
international journal.
- Hundreds of populations; use 15-
2.1 Primary Document: Data that
30% of sample size
is the most closely matches the study.
- Thousands of populations; use 10-
15% of sample size

Vol. 10 No. 2 July – December 2021 Page 77

 2.2 Secondary Document: Data 4. Data Analysis: Researchers
that has some inaccuracies from primary analyze data from primary document and
document. secondary document by both Content
3. Data Collection: All research Analysis and Statistical Analysis depending
data, researchers set criteria of data on variable types, variable measurement
collection including 1) Authenticity 2) scales, Type I error (Alpha, 𝛼), Type II
Credibility 3) Representativeness and 4) error (Beta, 𝛽), Power of Test: 1- 𝛽 and
Meaning. Effect Size by G*Power Ver.3.1.9.7.

Figure 1 The relationship of population size and derivation of sample size (Chanuan,
2020).

Research Methods and Research Results Yamane (1973) adjusted calculation

Sample Size Determination by formula to be more accurate; by increasing
calculation formula of Yamane compares of 𝜋 = population variance from
with Cochran Table. All 5 cases Cohen Dichotomous Variable equal to 0.50 and z =
compares with sample size by calculating z score at significance level 𝛼 (where z = 2
computer programs and Green. at 𝛼 = 0.05 and z = 3 at 𝛼 = 0.01) as the
1.Taro Yamane Formula (Yamane, 1973) following formula

N (𝑧)2 (𝜋)(1 − 𝜋)(𝑁)

𝑛= 𝑛=
1 + N𝑒 2 (𝑧)2 (𝜋)(1 − 𝜋) + (𝑁)(e)2
37,581
𝑛= = 395.78 𝑛
1 + 37,581x0.052 (2)2 (0.50)(1 − 0.50)(37,581)
=
≈ 396 (2)2 (0.50)(1 − 0.50) + (37,581)(0.05)2
Where n = sample size = 395.79 ≈ 396
N = population size = 37,581
From calculation formula of given
𝑒 = error (0.05) reliability level 95%
37,581 population; sample size equals to 396
or; 𝑒 = level of precision always set the
compares with Taro Yamane Table at
value of 0.05
Page 78 : APHEIT INTERNATIONAL JOURNAL
reliability level 95% (𝑒 = ±5%) and 𝑧 = 𝑧 value at reliability level or
various errors; sample size equal to 394–397 significance level.
which have a similar value. - Reliability level 95% or
2. Krejcie & Morgan Formula (Krejcie significance level 0.05; 𝑧 = 1.96
& Morgan, 1970) If the population size is - Reliability level 99% or
known significance level 0.01; 𝑧 = 2.58
3.2 If the population size is
unknown, the population proportion is
𝜒 2 𝑁𝑝(1 − 𝑝) unknown.
𝑛=
𝑒 2 (𝑁 − 1) + 𝜒 2 𝑝(1 − 𝑝)
𝑧2
𝑛=
𝑛 4𝑒 2
3.84137,5810.5(1 − 0.5) (1.96)2
= 𝑛= = 384.16
0.052 (37,581 − 1) + 3.8410.5(1 − 0.5) 4(0.05)2

36,087.16 𝑛 = sample size

𝑛= = 380.225 ≈ 380 𝑝 = the population proportions
93.95 + 0.96
𝑒 = acceptable sampling error (𝑒 =
n = sample size 0.05)
N = population size = 37,581 𝑧 = 𝑧 value at reliability level or
𝑒 = acceptable error of sample size significance level.
𝜒 2 = Chi-square 𝑑𝑓 = 1 and - Reliability level 95% or
reliability level 95% (𝑋 2 = 3.841) significance level 0.05; 𝑧 = 1.96
𝑝 = the population proportions - Reliability level 99% or
(Assumed to be 0.5) significance level 0.01; 𝑧 = 2.58
Compare calculation formula with 3.3 If the population size is
Krejcie & Morgan Table; sample size is 380 unknown and estimated population mean
which equals to the population proportion
𝜎 2𝑧2
and assumed to be 0.5; acceptable error 5% 𝑛=
and reliability level 95% but the advantage 𝑒2
by using table can calculate small and 10 or 152 𝑥 1.962
𝑛= = 35 𝑢𝑛𝑖𝑡
more sample size and population. 52
3. Cochran Formula (Cochran, 1977) 𝑛 = sample size (unit)
3.1 If the population size is 𝜎= standard deviation of the sample
(𝜎 = 15)
unknown but a lot, the population
𝑒 = acceptable sampling error ≈ (±5%)
proportion is known
[If 𝜎 is unknown, defined 𝑒 as % of
𝑝(1 − 𝑝)𝑧 2 𝜎 such as 8% of 𝜎 (𝑒 = 0.08𝜎) or 10% of
𝑛= 𝜎 (𝑒 = 0.10𝜎)]
𝑒2
(0.1)(1 − 0.1)(2.58)2 𝑧 = 𝑧 value at reliability level or
𝑛= = 240
(0.05)2 significance level.
𝑛 = sample size - Reliability level 95% or
𝑝 = the population proportion (𝑝 = 0.1) significance level 0.05; 𝑧 = 1.96
𝑒 = acceptable sampling error (𝑒 = - Reliability level 99% or
0.05) significance level 0.01; 𝑧 = 2.58
3.4 If the population size is known
and estimated the population proportion.

Vol. 10 No. 2 July – December 2021 Page 79

 𝑝(1 − 𝑝) not take into account in population size;
𝑛=
𝑒2 𝑝(1 − 𝑝) number of independent variables are only
2 +
𝑧 𝑁 known. Green had developed equation
0.5(1 − 0.5) (Harris, 1975) as n ≥ 50 + m and (Howell,
𝑛= = 322
0.052 0.5(1 − 0.5) 2002) noted that Green Formula is
+
1.962 2,000 appropriate for Regression Analysis more
than other formulas but not suitable for few
𝑛 = sample size predictors. For 5 predictors, the power of a
𝑁 = population size hypothesis test is 0.80
𝑒 = acceptable sampling error
𝑝 = the population proportions 𝑛 ≥ 50 + 8(𝑚)
𝑧 = 𝑧 value at reliability level or
significance level. where 𝑛 is sample size
- Reliability level 95% or 𝑚 is predictor or independent
significance level 0.05; 𝑧 = 1.96 variables If 𝑚 = 13, sample size estimated
- Reliability level 99% or
𝑛 ≥ 50 + 8(13) = 154
significance level 0.01; 𝑧 = 2.58
3.5 If the population size is known 5. Jacob Cohen Formula
and estimated population mean 5.1 Calculation of sample size by
group as follow:
𝑁𝑍 2 𝜎 2 𝑁0.05
𝑛= 𝑛= +1
(𝑁 − 1)𝑒 2 + 𝑍𝜎 2 400𝑓 2
400𝑥(1.96)2 𝑥 (15)2 1,096
𝑛= 𝑛= +1=7
(400 − 1)(5)2 + 1.96𝑥(15)2 400(0.70)2
= 33 𝑢𝑛𝑖𝑡
𝑛 = sample size (unit) Where 𝑁0.05 = derived from table
𝑁 = population size (unit) (Cohen, 1977: p.384); effect size from
𝜎
𝜎= standard deviation of the sample 𝑓 = 𝜎𝑚 = 0.70 One-way Analysis of
𝑒 = acceptable sampling error Variance test) and Power of test as defined
[If 𝜎 is unknown, defined 𝑒 as % of at significance level ( 𝛼) = 0.05 or 0.01 such
𝜎 such as 8% of 𝜎 (𝑒 = 0.08𝜎) or 10% of as significance level 0.05 and Power of test
𝜎 (𝑒 = 0.10𝜎)] 80% and u = 3 (Group 1) = 1096 unit;
𝑧 = 𝑧 value at reliability level or therefore, calculation of sample size is 7 unit
significance level. per group.
- Reliability level 95% or 5.2 Calculation of sample size for
significance level 0.05; 𝑧 = 1.96 Parametric Statistics by Regression Analysis
- Reliability level 99% or (Chua, 2006) as follow;
significance level 0.01; 𝑧 = 2.58 Calculation of sample size of Cohen;
independent variables k can be maximum of
4, Samuel B. Green Formula (Green, 10 ( u = 1 0 ), Power of Test = 0. 8 0 , Effect
1991) Size : 𝑓 2 = 0.15 where Type I error (Alpha,
Regression Analysis is calculation of 𝛼) = 0.05
sample size from Green Formula which is 𝜆
the representative of population; it suitable 𝑁= 2
𝑓
for research by Sample Survey in data
collection in order to test hypothesis by where 𝑓 2 : (Effect Size) from
Regression Analysis. The dominant equation (Cohen, J. 1992: pg.115)
characteristic of Green Formula which does

Page 80 : APHEIT INTERNATIONAL JOURNAL

 2
𝑅2 Results from calculation many
𝑓 = times; it found that 𝜆 value at v = 60 and v
1 − 𝑅2
= 120 will little decrease N, therefore the
substitute 𝑓 2 to the main equation, so appropriate value from calculation of
Calculation Formula of sample size of sample size is N = 116.
Cohen, 5.3 Calculation of sample size by
G*Power Program
𝜆 (1 − 𝑅 2 ) It can calculate sample size or power
𝑁=
𝑅2 test of many analytical statistics such as
exact statistics, t - tests, F - tests, Chi -
For a trail value of v = 120, 𝜆 = 17.4 square, Z tests, ANOVA, Correlation and
(from table 9.4.2, Cohen, 1988). Substitute 𝜆 Regression.
𝜆 Calculation of sample size by
to sample size equation (𝑁 = 𝑓2 ), therefore
G*Power Program in case of One-way
N = 17.4/0.15 = 116, and (v = N-u-1) = ANOVA; data input into computer program,
116-10-1 = 105 analytical statistics and other values
However, v value at v = 60 and v = including types of variables, variables scale,
120; N is the best value so Type I error (Alpha, 𝛼), Type II error (Beta,
1 1 𝛽), Power of Test: 1- 𝛽 and Effect Size as
𝑉𝐿 − 𝑉 follow (Cohen, 1988)
𝜆 = 𝜆𝐿 − ( 𝜆𝐿 − 𝜆𝑈 ) Estimated Sample Size for One-way
1 1
− ANOVA (Faul et al., 2009) as shown in the
𝑉𝐿 𝑉𝑈
figure 2, 3.
𝜆𝐿 = where v = 60 F- test for Group Effect
𝜆𝑈 = where v = 60 𝐻0 : delta = 0 versus 𝐻𝐴 : delta! = 0
𝑉𝐿 = v low Study Parameters:
𝑉𝑈 = v high Alpha = 0.0500
where v = 60, λ = 18.7, and v = 120, λ = Power = 0.8000
17.4. substitute λ to equation delta = 0.7014
N_g = 4
1 1
− m1 = 68.0000
𝜆 = [18.7 − 60 105 (18.7 − 17.4)]
1 1 m2 = 72.0000
60 − 120 m3 = 77.0000
= 17.58 m4 = 80.0000
Var_m = 21.1875
therefore Var_e = 43.0650
Estimated sample sizes:
𝜆 17.8
𝑁= = = 117 N = 28
𝑓 2 0.15 N per Group = 7

Vol. 10 No. 2 July – December 2021 Page 81

 Asst. Prof. Chanuan Uakarn, Ph.D.

Graph 𝛼, 𝛽 และ Critical Value

of each statistical test

1 Parameter Values

2
3

Input Parameters
Power
𝛼
Effect Size

Calculate sample size

Figure 2 Shows input parameters of various statistical test by G*Power Program in case of
One-way ANOVA.

Page 82 : APHEIT INTERNATIONAL JOURNAL

Figure 3 The relationship of total sample size and power of test: Power (1 − 𝛽).

Table 1 Input procedure as statistical values required for calculation of sample size by
G*Power (3.1.9.7). (Faul et al., 2009) as shown in the figure 4, 5 and 6.

Statistical Values  = 0.05, power (1-) = 0.80 Effect size = 0.30 and 𝜇 2 =0.088
Procedure Calculation of Sample Size
Type Type Type Data Input
Type 1 Type 2 Type 3
1 2 3
1 1 1 Test family F-Test F-Test F-Test
ANOVA: Fixed
ANOVA: Fixed ANOVA: Fixed effect,
Statistics
2 2 2 effect, omnibus, effect, omnibus, Special, main
test
one-way one-way effects and
interactions
A priori A priori A priori
Type of
compute compute compute
3 3 3 power
required sample required sample required sample
analysis
size-given 𝛼, size-given 𝛼, size-given 𝛼,

Vol. 10 No. 2 July – December 2021 Page 83

 power, and power, and power, and
effect size effect size effect size
Input
− 4 4 − Determine=≫ Determine=≫
Parameter
Select Effect size from
− 5 − − −
Procedure means
Number of
4 6 5 3 3 3
groups
− SD 𝜎 within
− 7 − 18.25 −
each group
− 8 − Mean − 45.7, 43.5, 32.6 −
− 9 − Equal n − 50 −
− 10 6 Size − 50 Direct
Total Partial
− 11 7 − 150
sample size 𝜇 2 =0.088
Effect size Effect size
− 12 8 Calculate −
f=0.3138471 f=0.31063
Calculate
and transfer Effect size Effect size Effect size
5 13 9
to main f=0.30 (Cohen) f=0.3138471 f=0.31063
window
Total sample Total sample Total sample
6 14 10 Calculate
size 111 size 102 size 178
Critical F 3.08039 3.08824 1.88578

Page 84 : APHEIT INTERNATIONAL JOURNAL

 Asst. Prof. Chanuan Uakarn, Ph.D.

Calculation of
sample size
type 1

1
2

4
Critical F

Sample Size = 111

Figure 4 Shows calculation of sample size type 1.

Vol. 10 No. 2 July – December 2021 Page 85

 Asst. Prof. Chanuan Uakarn, Ph.D.

Calculation of
sample size
type 2

5
1
2 6
7
3
9
8
4

Critical F 10

12 11
Calculate and transfer
Sample Size = 102
to main window 13

14 13

Figure 5 Shows calculation of sample size type 2.

Page 86 : APHEIT INTERNATIONAL JOURNAL

 Asst. Prof. Chanuan Uakarn, Ph.D.

Calculation of
sample size
type 3
1
2
3

6
5
7
Calculate and transfer Sample Size = 178 8
to main window 9

10 9

Figure 6 Shows calculation of sample size type 3.

Discussion and recommendation Conclusion and recommendation

According to Calculation of Sample Sample Size Determination of Yamane and
Size by formulas and tables of academic officer Krejcie & Morgan are suitable for Survey
together with input various statistical values, Research and Finite Population; but not
these are in order to receive the real value of suitable for Experiment Research. In case the
representatives of population and correct population size is unknown but the number is
sample size. Results from similar experiments high, population proportion is known or the
such as Taro Yamane and Krejcie & Morgan population size is unknown, population
However, the sample size is still high. In terms proportion is unknown or the population size is
of Green and Cohen, the sample size is smaller, unknown and estimated population mean or the
or not relatively not close to Cochran these are population size is known and estimated the
depending on size or number of inputs, population proportion Cochran Formula is
statistical values, acceptable error, power of suitable for all these. Green Formula is easy
test, effect size, population size, variable size. and suitable for Regression Analysis; number
The most important is the selection of statistical of independent variables are only known.
methods for analyzing Inferential Statistics Cohen Formula is another choice; the
with parameters such as t-Test, F-Test, z-Test, calculation depends on ( 𝛼), Power of Test,
ANOVA, Regression Analysis or Inferential Effect Size, Type I or II error and correspond
Statistics without parameters for instance with calculation by G*Power.
correlation and Chi-Square.

Vol. 10 No. 2 July – December 2021 Page 87

Reference
Chua Lee Chuan. (2006). Sample Size Estimation Using Krejcie and Morgan and Cohen
Statistical Power Analysis: A Comparison. Jurnal Penyelidikan IPBL, Jilid 7.
Cochran, W. G. (1977). Sampling techniques. 3rd Ed. New York: John Wiley & Sons.
Cohen, J. (1977). Statistical Power Analysis for Behavioral Sciences. 1st Ed. New York:
Academic Press. StataCorp. 2003. STATA Reference Manual Release 10. Texas: Stata
Press.
Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences, 2nd Ed. Hillsdale, New
Jersey: Lawrence Erlbaum Associates.
Cohen, J. (1992). Quantitative Methods in Psychology: Power Premier. Psychological
Bulletin; July1992; Vol.112, No.1; pp.155-159
Chanuan Uakarn. (2020). Advanced Quantitative Research Methods in Public Policy. Kasem
Bundit University: Public Policy and management, Lecture document 2-2020.
Faul, F., Erdfelder, E., Lang, A.-G., and Buchner, A. (2007). G*Power 3: A flexible statistical power
analysis program for the social, behavioral, and biomedical sciences. Behavior Research
Methods, 39, 175-191.
Faul, F., Erdfelder, E., Buchner, A., and Lang, A.-G. (2009). Statistical power analyses using
G*Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods,
41, 1149-1160.
Krejcie, R.V., and Morgan, D.W. (1970) Determining Sample Size for Research Activities.
Educational and Psychological Measurement. 30, 607 – 610.
Samuel B. Green, (1991). How Many Subjects Does It Take To Do A Regression Analysis? Jul
1;26(3), 499-510.
Sedlack R. G., and Stanley J. (1992). Social Research: Theory and Methods. 1st Ed., Indiana
University: Allyn and Bacon.
Yamane, Taro. (1973), Statistics: An Introductory Analysis. London: John Weather Hill, Inc.

Page 88 : APHEIT INTERNATIONAL JOURNAL