• SAMPLE SIZE ESTIMATION USING KREJCIE AND MORGAN AND


  •   
  • FileName: chua06.pdf [read-online]
    • Abstract: SAMPLE SIZE ESTIMATION USING KREJCIE AND MORGAN ANDCOHEN STATISTICAL POWER ANALYSIS: A COMPARISONChua Lee ChuanJabatan PenyelidikanABSTRACTIn most situations, researchers do not have access to an entire

Download the ebook

SAMPLE SIZE ESTIMATION USING KREJCIE AND MORGAN AND
COHEN STATISTICAL POWER ANALYSIS: A COMPARISON
Chua Lee Chuan
Jabatan Penyelidikan
ABSTRACT
In most situations, researchers do not have access to an entire
statistical population of interest partly because it is too expensive
and time consuming to cover a large population or due to the
difficulty to get the cooperation from the entire population to
participate in the study. As a result, researchers normally resort to
making important decisions about a population based on a
representative sample. Hence, estimating an appropriate sampling
size is a very important aspect of a research design to allow the
researcher to make inferences from the sample statistics to the
statistical population. The power of a sample survey lies in the
ability to estimate an appropriate sample size to obtain the
necessary data to describe the characteristics of the population.
With that as the rationale, this article was written to make
comparison between two commonly used approaches in estimating
sampling size: Krejcie and Morgan and Cohen Statistical Power
Analysis. It also highlights the significance of using Cohen’s
formula over Krejcie and Morgan’s for higher accuracy to base
decisions on research findings with confidence.
INTRODUCTION
For most studies that require data from a wide and diverse population size, rarely do
researchers cover the whole population. The normal practice is to draw a sample
from the target population. Salant and Dillman (1994) defined a sample as a set of
respondents selected from a larger population for the purpose of a survey. The main
reason to sample is to save time and money. Furthermore, it is generally not
necessary to study all possible cases to understand the phenomenon under
consideration (Ary, Jacobs, & Razavieh, 1996). The most important thing taken into
consideration is that the sample drawn from the population must be representative
so that it allowed the researcher to make inferences or generalisation from the
sample statistics to the population understudied (Maleske, 1995). If the sample size
is too low, it lacks precision to provide reliable answers to research questions
investigated. If the sample size is too large, time and resources could be wasted
often for minimal gain. Therefore, the power of a sample survey actually lied in the
Jurnal Penyelidikan IPBL, Jilid 7, 2006___________________________________________
79
ability to obtain the necessary information from a relatively few respondents to
describe the characteristics of the entire population.
DETERMINING THE SAMPLE SIZE
For any research, the sample size of any study must be determined during the
designing stage of the study. However, before determining the size of the sample
that needed to be drawn from the population, a few factors must be taken into
consideration. According to Salant and Dillman (1994), the size of the sample is
determined by four factors: (1) how much sampling error can be tolerated; (2)
population size; (3) how varied the population is with respect to the characteristics
of interest; and (4) the smallest subgroup within the sample for which estimates are
needed.
Using the above methods as a guideline, the following section aims to compare two
approaches in determining the sample size of a population of 500 people using (a)
Krejcie and Morgan (1970) and (b) Cohen Statistical Power Analysis.
Krejcie and Morgan
Estimation of sample size in research using Krejcie and Morgan is a commonly
employed method. Krejcie and Morgan (1970) used the following formula to
determine sampling size:
S = X2NP (1-P)/ d2 (N-1) + X2P(1-P)
S = required sample size
X2 = the table value of chi-square for one degree of freedom at the desired
confidence level
N = the population size
P = the population proportion (assumed to be .50 since this would
provide the maximum sample size)
d = the degree of accuracy expressed as a proportion (.05)
Based on Krejcie and Morgan’s (1970) table for determining sample size, for a
given population of 500, a sample size of 217 would be needed to represent a cross-
section of the population. However, it is important for a researcher to consider
whether the sample size is adequate to provide enough accuracy to base decisions
on the findings with confidence. Therefore, in order to find out if the sample size
recommended by Krejcie and Morgan (1970) is sufficient, the next section aims to
illustrate the estimation of sampling size using Cohen’s (1988) statistical power
analysis.
Jurnal Penyelidikan IPBL, Jilid 7, 2006___________________________________________
80
Cohen Statistical Power Analysis
According to Cappelleri and Darlington, (1994), Cohen Statistical Power Analysis
is one of the most popular approaches in the behavioural sciences in calculating the
required sampling size. According to Cohen (1998), in order to perform a statistical
power analysis, five factors need to be taken into consideration:
1. significance level or criterion
2. effect size
3. desired power
4. estimated variance
5. sample size
Cohen (1988) statistical power analysis exploits the relationships among the five
factors involved in statistical inferences. For any statistical model, these
relationships are such that each is a function of the other four. Taking that into
consideration, it means that if sample size is to be determined, it can be estimated
for any given statistical test by specifying values for the other four factors: (1)
significance level, (2) effect size, (3) desired power and (4) estimated variance.
When Cohen’s statistical power analysis is used to determine the sample size, the
objective of the analysis is to calculate an adequate sampling size so as to optimise
as opposed to maximising sampling effort within the constraint of time and money.
Optimising sampling efforts will avoid situations where lack of subjects is
considered giving rise to inconclusive inference-making. Contrary, maximising
sampling efforts occur when the collection of data goes beyond the required level to
achieving significant results, thereby, limited resources are wasted.
In order to determine an adequate sample size, the values of significance level,
effect size, power and estimated variance have to be pre-determined.
The statistical level of significance for most studies in the teaching field is often
fixed at alpha = .05. Alpha is the probability of wrongly rejecting the null
hypothesis, thus committing Type I error. Assigning a less stringent alpha would
increase the risk of false rejection or ‘crying wolf’ (Eagle, 1999), casting doubts on
the validity of the results. However, if the alpha is too conservative, evidence from
the findings might fail to reject the null hypothesis in the presence of substantial
population effect. Therefore, setting the alpha at .05, is considered the most
conventional level of significance, which is normally used in the field of education.
(Ary, et al., 1996).
The next factor to be determined is the effect size. Effect size generally means the
degree to which the phenomenon is present in the population or the degree to which
the null hypothesis is false (Cohen, 1988). It essentially measures the distance or
discrepancy between the null hypothesis and a specified value of the alternative
hypothesis. Each statistical test has its own effect size index. All the indexes are
scale free and continuous ranging from zero upwards (Cohen, 1992). For any
Jurnal Penyelidikan IPBL, Jilid 7, 2006___________________________________________
81
statistical test, the null hypothesis has an effect size of zero. For example, in using
the product-moment correlation to test a sample for significance, the effect size
index is r, and H0 posits that r = 0. For multiple regression, the effect size index is f2
and H0 posits that f2 = 0.
Effect size can be measured using raw values or standardised values. Cohen has
standardised effect sizes into small, medium and large values depending on the type
of statistical analyses employed. The effect sizes to test the significance of product-
moment correlation coefficient, r, are, .10, .30, and .50, for small, medium and large
respectively. For regression analysis, the effect size index, f2 for small, medium and
large effect sizes are f2 = .02, .15, and .35 respectively. The smaller the effect size,
the more difficult it would be to detect the degree of deviation of the null hypothesis
in actual units of response. Cohen (1992) proposed that a medium effect size is
desirable as it would be able to approximate the average size of observed effects in
various fields. Cohen (1992) also argued that a medium effect size could represent
an effect that would likely be “visible to the naked eye of a careful observer”
(p156).
Next to determine is the statistical power. The power of a statistical test is defined
as the probability that a statistical significance test will lead to the rejection of the
null hypothesis for a specified value of an alternative hypothesis (Cohen, 1988).
Power analysis has the ability to reject the null hypothesis in favour of the
alternative when there is sufficient evidence from a collected sample that a value of
a parameter from the population of interest is different from the hypothesised value
(High, 2000). Putting it simply, it is the probability of correctly rejecting the null
hypothesis given that the alternative hypothesis is true.
In statistical parlance, power is expressed as 1-β, where β is the probability of
wrongly accepting the null hypothesis when it is actually false or failure to reject
null hypothesis that is false. This is known as committing Type II error. The value
can range between zero to one.
According to Thomas and Juanes (1996), power analysis is a critical component in
designing experiments and testing results. However, computing power for any
specific study can be a difficult task. High (2000) argued that when low power is
used in a study, the risk of committing Type II error is higher, that is, there is little
chance of detecting a significant effect, which can give rise to an indecisive result.
Stating it differently, the effect is there but the power is too low to detect it.
However, if the power is set too high, a small difference in the effect is detectable,
which means that the results are significant, but the size of the effect is not practical
or of little value. In addition, a larger power would result in a demand for N that is
likely to exceed the resources of the researcher (Cohen, 1992). To avoid these
problems, Cohen (1992) suggested fixing the power at .80 (β = .20), which is also a
convention proposed for general use. However, this value is not fixed. It can be
adjusted depending on the type of test, sample size, effect size as well as the
sampling variation.
Jurnal Penyelidikan IPBL, Jilid 7, 2006___________________________________________
82
The fourth and last factor to determine is standard deviation, which is often used for
estimating the variation in the response of interest. This value can be obtained,
either from previous studies or pilot studies. However, when standardised measures
are dimensionless quantities, the sampling variance is already implicitly
incorporated. Such standardised measures include the d-values or correlation
coefficients and as such the value of variance is not required (Thomas & Krebs,
1997). Therefore if study aims to look at the correlation of variables, this value is
not needed for calculating the sample size of the study.
Using the factors mentioned above to estimate sample size, the next section aims to
illustrate the use of the Cohen Statistical Power Analysis to calculate an adequate
sample size. However, before the sample size is estimated, researchers need to
predetermined factors pertaining to alpha size, effect size and power. Additionally,
it is also important for researchers to know the underlying objectives of the study
and how the data will be analyzed to achieve the objectives. This is because, the
sampling size varies according to the type of statistical tests performed on the data
gathered.
For instance, the factors pre-determined in order to estimate an adequate sample
size for a study are, the alpha level is set at .05, the effect size is medium and the
power is set at .80. For illustrative purposes, two statistical tests will be used to
analyse the data of a study, such as, Pearson Product Moment Correlation and
Multiple regression analysis.
Using the predetermined values and the two statistical tests as guidelines, the next
section will illustrate on how to calculate a suitable sample size.
ILLUSTRATIVE EXAMPLES
The sample size for Pearson Product Moment Correlation Analysis and Multiple
Regression Analysis can be easily determined using Cohen statistical power
analysis.
Pearson Product Moment Correlation Analysis
If a study aims to find out the degree of the relationship (non-directional) between a
dependent variable and ten independent variables, with a predetermined effect size
of r = .30 (medium), a significant alpha = .05 and a statistical power of .80, the
desired sample size to test these relationships as indicated in Table 3.4.1 is 85
(Cohen, 1992). This means that 85 respondents are sufficient to perform this
statistical analysis.
Jurnal Penyelidikan IPBL, Jilid 7, 2006___________________________________________
83
Regression Analysis
If the study also aims to investigate the contribution of each of the ten predictor
variables towards the variance of a dependent variable. This investigation required
the use of multiple regression analysis. To estimate the sample size for the
regression analysis segment of the research, Cohen’s formula takes into
consideration the number of k independent variables used in the analysis.
Calculation can performed on the maximum of 10 independent variables (u = 10).
With the specified power of .80, a medium effect size of f 2 = .15, a significant alpha
of .05, Cohen’s statistical power analysis formula to calculate the sample size
needed for this analysis is
N = λ / f2
This formula required the determination of unknown lambda value, λ, which is then
needed to find the necessary sample size, N. However, the lambda value depended
on the degree of freedom of the denominator of the F ratio, v.
v=N–u–1
To account for this problem, a trial value of v is taken from Table 9.4.2 (Cohen,
1988) to obtain the lambda value which is needed to compute N. If the computed N
implied v substantially differed from the trail value, the computation of the new v
value has to be used.
For a trail value of v = 120, λ = 17.4 (Table 9.4.2, Cohen, 1988). Substituting λ into
the sample size formula ( N = λ / f2 ), gives N = 17.4/.15 = 116, which implied that
v = 116-10-1 = 105.
However a more accurate value for N required reiteration by interpolating between
lambda values for v = 60 and v = 120 where
1/vL –1/v
λ = λL – (λL - λU)
1/vL –1/vU
λL = lambda value when v = 60
λU = lambda value when v = 120
vL = lower v value
vU = upper v value
When v = 60, λ = 18.7, when v = 120, λ = 17.4. Substituting λ into the formula, the
exact λ value =
1/60 –1/105
λ = 18.7 – (18.7 - 17.4) = 17.58
1/60 –1/120
Jurnal Penyelidikan IPBL, Jilid 7, 2006___________________________________________
84
Therefore N = 17.58/.15 = 117. The result showed that reiteration by interpolating
for λ between v = 60 and v = 120 did not significantly change the previous value.
Therefore, no further iteration is necessary and the originally computed N = 116 is
maintained.
COMPARISON OF ESTIMATED SAMPLING SIZE
Based on the above calculation using Krejcie and Morgan (1970), the estimated
sampling size for a population of 500 is 217. However, the estimated sampling size
calculated using Cohen (1992) differs according to the type of statistical tests
employed by the researcher. The sample size that is required for a correlational
study is 85 while a multiple regression analysis requires 116. This indicates that the
sampling size can range from a minimum of 85 for performing correlation analysis
to a maximum of 217 as recommended by Krejcie and Morgan (1970).
Estimating an adequate number of respondents is critical to the success of a
research. According to High (2000), the size of the study sample is critical to
producing meaningful results. When there were too few subjects, it might be
difficult to detect the effect or phenomenon understudied, thus providing
inconclusive inference-making. On the other hand, if there were too many subjects,
even trivially small effect could be detected, but the findings would be of
insignificant value, wasting valuable time and resources.
Most studies are conducted using Cohen’s (1988) statistical power analysis as the
guideline for estimating the desired sample size. A few reasons justified the use of
this analysis. First, Cohen is not only concerned about the magnitude with regards
to the statistical test results and its accompanying ρ value (as most researchers are)
but also the existence of the phenomenon understudied by considering additional
factors such as population effect size and the statistical power. In most research,
significance testing is heavily preferred to confidence interval estimation (Cohen,
1992). They failed to consider the importance of effect size and the statistical
power, which has been established in the preceding section. Considering all these
factors as suggested by Cohen (1988) would lead to more meaningful results than
results that have been inferred from the observed p-value. Furthermore, lacking of
controversies among methodologies on the importance of Cohen’s (1988) statistical
power analysis and the availability of ample resources for estimating sample sizes in
research designs using power analysis, this analysis has achieved high reliability for
determining an appropriate sample size.
Based on the above justifications, the sample size calculated using the formula
derived from Cohen’s Statistical power analysis would be more meaningful and
acceptable. A sample size of 217 as recommended by Krejcie and Morgan (1970)
would be too large a number. Conducting a study, which involved too many
subjects than what is deemed necessary would mean that valuable time and
resources were not used efficiently and economically.
Jurnal Penyelidikan IPBL, Jilid 7, 2006___________________________________________
85
THE FINAL SAMPLE SIZE
The sample size can be increased to N = 120, slightly more than the recommended
size. This number can be rounded up (from 116) to allow the researcher to execute
Cohen’s (1988) table for further analysis of the power level. A sample size of 120
would be sufficient to answer research objectives using, both, correlation analysis
and multiple regression analysis.
With a new sample size of N = 120, it is necessary to estimate the power value for
the two statistical analyses. This is essentially important to ensure that the
predetermined power value of .80 be achieved, which according to Cohen (1998), is
the probability that a statistical significance test can gather enough evidence to
correctly reject the null hypothesis in favour of the alternative hypothesis. Since the
value of power varied with the type of statistical analysis performed, still
maintaining the predetermined statistical criterion of .05 and medium effect size (r =
.30), the power value for correlation analysis is increased from .80 to .92 (Cohen,
1988: Table 3.3.5).
As for multiple regression analysis, the calculation of power also required the
values of the predetermined factors. Based on the sample size of 120 and the
predetermined statistical criterion α = .05, medium effect size, f2 = .15, the
calculated power for multiple regression analysis is .807.
Based on the calculated power values for the two statistical analyses, with a sample
size of 120, the values ranged from .80 to .92. These reported values achieved the
minimum proposed value of .80 from Cohen (1988).
Considering the seriousness of type I and type II errors and the cost of obtaining
data, this sample size is adequate and manageable. A sample size of 120 is adequate
as it has the ability to detect an effect at the desired power equal to a minimum of .
80 or even larger. Reducing the sample size would reduce the power value to
below .80, which would be undesirable. As a result, 120 respondents can be
randomly selected from the target population to participate in this study.
REFERENCES
Ary, D., Jacobs, L. C., & Razavieh, A. (1996). Introduction to research in
education. Orlando, Florida: Harcourt Brace College Publishers.
Cappelleri, J. C., & Darlington, R. B. (1994). The power analysis of cutoff-based
randomized clinical trials. Evaluation Review, 18(2).
Cohen, J. (1988). Statistical power analysis for the behavioral sciences. Second
Edition. Hillsdale, NJ: Lawrence Erlbaum Associates, Publishers.
Cohen, J. (1992). Quantitative methods in psychology: A power primer.
Psychological Bulletin, 112(1), 155-159.
Jurnal Penyelidikan IPBL, Jilid 7, 2006___________________________________________
86
Eagle, P. C. (1999). Power: A Primer and other take home messages.
http://www.mpl-pwrc.usgs.gov/powcase/primer.html. Accessed on 1
December 2002.
High, R. (2000). Important factors in designing statistical power analysis studies.
Computing News, Summer issue, 14-15.
Krejcie, R. V., & Morgan, D. W. (1970). Determining sample size for research
activities. Educational and Psychological Measurement, 30, 607-610.
Maleske, R. T. (1995). Foundations for gathering and interpreting behavioral data.
Pacific Grove, CA: Brooks/Cole Publishing Company.
Salant, P., & Dillman, D. A. (1994). How to conduct your own survey. New York:
John Wiley & Sons, Inc.
Thomas, L., & Juanes, F. (1996). The importance of statistical power analysis: An
example from animal behaviour. Animal Behaviour, 52, 856-859.
Thomas, L., & Krebs, C. J. (1997). A review of statistical power analysis software.
Bulletin of the Ecological Society of America, 78(2), 126-139.
Jurnal Penyelidikan IPBL, Jilid 7, 2006___________________________________________


Use: 0.0554