Sample Size Determination: Maj. Tun Tun Win

Sample size
determination
Maj. Tun Tun Win

Lecturer
Dept. of Preventive and Social Medicine
Importance of Sample Size
calculation
• Scientific reasons
• Ethical reasons
• Economic reasons
3/7/22 dr.ttunwin@gmail.com 2
From a scientific perspective:
• Can’t be sure we’ve made right decision

regarding the effect of the intervention
• However, we want enough subjects enrolled
to adequately address study question to
feel comfortable that we’ve reached correct
conclusion
From an ethical perspective:
Too few subjects:
• Cannot adequately address study question.

The time, discomfort and risk to subjects
have served no purpose.
• May conclude no effect of an intervention

that is beneficial. Current and future
subjects may not benefit from new
intervention based on current (inconclusive)
study.
Too many subjects:
• Too many subjects unnecessarily

exposed to risk. Should enroll only
enough patients to answer study
question, to minimize the discomfort
and risk subjects may be exposed to.
Economic Reasons
• Undersized study is a waste of resources due to

its inability to yield useful results
• Oversized study may result in statistically

significant result with doubtful clinical importance
leading to waste of resources (Cardiac Studies)
First: Factors to Consider for Sample Size
• objectives
• study design
• resources and schedule of the study
Important questions in sample size
estimation
What is the key outcome of interest which is to be evaluated
statistically?
Cured/Not Cured, laboratory parameters, Success rate, etc.
How will the key outcome be measured?

Categorical or Numerical
What kind of study design does one have?

Descriptive (Parameter estimation), Analytic (Hypothesis testing)
A priori information about parameters of interest
Important questions in sample size estimation (2)
• A priori information about parameters of
interest
– Literature Review
From previous report, it was shown that cure rate of Drug
A = 70%
– Pilot Study
A pilot survey from 30 bottles of drinking water in the market
shows that there are E. coli in 5 bottles.
– 3Expert Opinion
out of 5 experts say that about 10% of workers in the XXX
factory have health problem related to toxic chemicals.
Second: Questions to Ask
• What is the null hypothesis?
• What is the desired level of significance? (α level)
• What is the power? (1- ß)
Hypothesis Testing
Null Hypothesis
• There is no difference between the groups
• Example. Survival rate in patients receiving treatment A is the
same as the survival rate in the group receiving treatment B.
Alternative Hypothesis
• There is a difference between the groups
• Example. Survival rate in patients receiving treatment A is better
than the survival rate in the group receiving treatment B.
Factors in sample size estimation
• Type I & Type II errors
Ho: 1 = 2 Reality/Truth
Ho is True Ho is False
CORRECT Type II ERROR
Do not
reject Ho Confidence: 1 -  
Decision
.99, .95 .10, .20

A B
Type I ERROR C D CORRECT
Reject Ho  Power : 1 - 
.01, .05 .90, .80
Type I and II Errors
• Type I error ():
– The probability that we reject Ho given that it is true
– The probability that we find an association between 2 factors
when, in truth, one does not exist.
• Type II error ():
– The probability that we failed to reject H0 when it is false
– The probability that we find no association between 2 factors
when, in truth, one does exist.
Confidence Level (1 -a)
• The probability of accepting a null hypothesis when it

is true is often called the confidence level (1- α ) of
the test
95% confidence level or  = 0.05
99% confidence level or  = 0.01

(need larger sample size)
Power (1- b)
• The chances of correctly rejecting a Null hypothesis
when it is indeed false is called the statistical
power (1- ß ) of the test of the hypothesis
– Typically chosen- (power) 0.8 – 0.95

– To obtain higher power - the larger sample size must
be obtained.
Sample Size and Power
• Too small ---> limit the conclusions

• Too large --> unnecessary time, money & men
--> unethically over exposure of subjects into
inferior treatment
• Compute smallest sample size to achieve the objectives

• To determine statistical power provided by a specified
sample size
Third: Factors in sample size estimation
• Effect size or Precision (d or Ԑ )
• Confidence level (in parameter estimation) / Tail

of the test (in hypothesis testing)
Type I error (α, in parameter estimation) / Type

I (α) & Type II (β) errors
(in hypothesis testing)
Sample size
1) Sample size to estimate P with d or ε
2) Sample size to estimate μ with d or ε
3) SS to estimate the risk difference
between two proportions within d
4) SS for two sample test of proportions
(Hypothesis testing)
5) SS estimating the difference between two
means
Sample size cont:
6) SS hypothesis testing for two population
means
7) SS for a hypothesis test of the Odds Ratio
8) SS to estimate the odds ratio
9) SS for a hypothesis test of the RR
10) SS to estimate the RR
11) SS for equivalence study
(1) Sample size to estimate P with d or ε
To study the prevalence of Hypertension among

military personnel in Mingalardon cantonment
• Previous study 20% of above 50 years were hypertensive

• Absolute precision ( 15% -25 % ) = 5%
• 95% Confidence level
• Sample size = n = ????
see table or computer
Table for One Proportion Study
(prevalence study)
Expected prevalence
Precision 5% 10% 15% 20% 25% 30% 35% 40% 50%
5% 73 138 196 246 289 323 350 369 380
6% 51 96 136 171 201 224 243 56 264
7% 37 71 100 125 147 165 178 188 194
8% 29 54 77 96 113 126 137 144 149
9% 23 43 60 76 89 100 108 114 117
10% 18 35 49 61 72 81 87 92 95
Sample size = n = z 2 ( p x q) /d2
p = 20% = 0.20
q (1-p) = 0.8
z = 1.96 ( at 95 % Confidence level )
d = 5 % = 0.05 (5 % precision)
n = z 2 ( p x q) /d2
n = 1.962(0.2 x 0.8) / 0.052
n = 3.841 x 0.16 / 0.0025
n = 245.86 = 246dr.ttunwin@gmail.com
3/7/22 22
(2) Sample size to estimate μ with d or ε
To study the average blood loss among elective LSCS
patients in No.(1) DSOGCH
n = Z 2 σ2
d2
n = Minimum required sample size
For 95% confidence level, z = 1.96
σ = standard deviation
d = Absolute precision
(3) Two Proportion (Estimation)
• Estimating the difference between two
population proportion with specified absolute
precision
n = z2 [ p1 (1-p1)+p2 (1- p2)]
d2
To study the proportion of hypertension difference between infantry units
and corps
(4) Two proportions (Hypothesis testing)
• Hypothesis test for two population proportion
To study the proportion of ICU utilization among

COVID-19 patients with or without comorbidities
n = [z1-α/2 √2p- (1-p-) + z1-β √p1(1-p1)+p2(1-p2)]2
(
p1 – p2)2
p- = p 1 + p 2
2
Two Proportions
• p1 = 20%
• p2 = 40%
•  = 0.05…….95% confidence interval
•  = 0.20…….(1- ) = 80%
• r =1 ( ratio of two group)
• Sample size = n = ???
• n = 91
(5) Two means (Estimation)
n = Z21-α/2 (2σ2)
d2
e.g. Nutritionist wish to estimate the difference in caloric intake at lunch
between children in a school offering a hot school lunch program and
children in a school which does not. From other nutrition studies, they
estimate that the standard deviation in caloric intake among elementary
school children is 75 calories and they wish to make their estimate to
within 20 calories or the future differences with 95% CI.
n = 1.962 (2* 752)
202
= 108.05
(6) Two Means (Hypothesis)
• To determine the effect of nutritional supplement vs.

vitamin A supplement over body weight of children
 Nutrition programme vs Vitamin A supplement
 Outcome variable weight (kg)
 To detect the differences of two means value of
weight of two groups of children.
n = (α+β)2 (σ12+ σ22)
(μ1- μ2)2
For 95% confidence level,
α = 0.05, z = 1.96
Power 80%,
β = 0.2, power = 1- β = 0.84
σ1= standard deviation of population 1
σ2= standard deviation of population 2
- μ1= Mean value of population 1
- μ2= Mean value of population 2
(7) Case control study
• Hypothesis tests for an odds ratio
• n = [z1-α/2 √2 p2(1-p2) + z1-β √p1(1-p1)+p2(1-p2)]2
(
p1 – p2)2
p1 = a/ a+b
P2 = c/c+d
In this case, probability of exposure given no
disease (p2) is approximated by the overall
exposure rate.
(8) Case control study
• Estimating an odds ratio with specified
relative precision
• Confidence level ?????
• Relative precision?????
p1 = a/ a+b
p2 = c/ c+d
OR ≥ 1, p2 and OR use for table.
OR < 1, p1 and 1/OR use for table.
(9) Cohort study
• Hypothesis tests for a relative risk
n = [z1-α/2 √2p- (1-p-) + z1-β √p1(1-p1)+p2(1-p2)]2
(
p1 – p2)2
p- = p1 + p2
2
p1 = a/ a+b
p3/7/22
2 = c/ c+d
dr.ttunwin@gmail.com 33
(10) Cohort study
• Estimating a relative risk with specified
relative precision
• Confidence level ?????
• Relative precision?????
p1 = a/ a+b
p2 = c/ c+d
RR ≥ 1, p2 and RR use for table.
RR < 1, p1 and 1/RR use for table.
(11) Sample size for equivalence study
Sample size is not
a magic number
dr.ttunwin@gmail.com
3/7/22 38

Sample Size Determination: Maj. Tun Tun Win

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Sample Size Determination: Maj. Tun Tun Win

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Sample Size Determination: Maj. Tun Tun Win

Uploaded by

Copyright:

Available Formats

Sample size

Maj. Tun Tun Win

• Can’t be sure we’ve made right decision

Too few subjects:

• Cannot adequately address study question.

• May conclude no effect of an intervention

• Too many subjects unnecessarily

• Undersized study is a waste of resources due to

• Oversized study may result in statistically

• resources and schedule of the study

How will the key outcome be measured?

What kind of study design does one have?

A priori information about parameters of interest

• What is the null hypothesis?

• What is the desired level of significance? (α level)

• What is the power? (1- ß)

than the survival rate in the group receiving treatment B.

.99, .95 .10, .20

• The probability of accepting a null hypothesis when it

95% confidence level or  = 0.05

99% confidence level or  = 0.01

– Typically chosen- (power) 0.8 – 0.95

• Too small ---> limit the conclusions

• Compute smallest sample size to achieve the objectives

• Effect size or Precision (d or Ԑ )

• Confidence level (in parameter estimation) / Tail

Type I error (α, in parameter estimation) / Type

To study the prevalence of Hypertension among

• Previous study 20% of above 50 years were hypertensive

To study the proportion of hypertension difference between infantry units

• Hypothesis test for two population proportion

To study the proportion of ICU utilization among

n = [z1-α/2 √2p- (1-p-) + z1-β √p1(1-p1)+p2(1-p2)]2

• To determine the effect of nutritional supplement vs.

n = [z1-α/2 √2p- (1-p-) + z1-β √p1(1-p1)+p2(1-p2)]2

You might also like