Scribd
Upload
51 views3 pages

Sample Size Formula - Lab CRP VI

This document provides formulas and guidance for calculating minimum sample sizes for different types of study designs, including descriptive studies of dichotomous and continuous variables, analytical studies of unpaired and paired dichotomous and continuous variables, and correlation studies. It includes notes on accounting for non-response, estimating proportions or standard deviations when pilot data is unavailable, and how to handle numeric or ordinal outcome variables.

Uploaded by

syarief hidayatullah
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
51 views3 pages

Sample Size Formula - Lab CRP VI

This document provides formulas and guidance for calculating minimum sample sizes for different types of study designs, including descriptive studies of dichotomous and continuous variables, analytical studies of unpaired and paired dichotomous and continuous variables, and correlation studies. It includes notes on accounting for non-response, estimating proportions or standard deviations when pilot data is unavailable, and how to handle numeric or ordinal outcome variables.

Uploaded by

syarief hidayatullah
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Descriptive study of a dichotomous variable

2
𝑍1−𝛼 2 𝑝𝑞
𝑛=
𝑑2

n = minimum sample size needed (rounded up)


Z1 – /2 = Z score for 1 – /2
p = proportion of outcome variable from previous study or pilot study
q =1–p
d = precision
make sure npq  5

Descriptive study of a continuous variable

2
𝑍1−𝛼 2 𝑠
𝑛=
𝑑

n = minimum sample size needed (rounded up)


Z1 – /2 = Z score for 1 – /2
s = standard deviation outcome variable from previous study or pilot study
d = precision

Analytical study of an unpaired-dichotomous variable (cross-sectional or cohort study)

Outcome/disease (+) Outcome/disease (-) Total


Predictor/exposure (+) a b a+b
Predictor/exposure (-) c d c+d
Total a+c b+d a+b+c+d

2
𝑍1−𝛼 2 2𝑝𝑞 +𝑍1−𝛽 𝑝 1 𝑞 1 +𝑝 2 𝑞 2
𝑛1 = 𝑛2 = for two-sided
𝑝1 − 𝑝2 2

2
𝑍1−𝛼 2𝑝𝑞 +𝑍1−𝛽 𝑝 1 𝑞 1 +𝑝 2 𝑞 2
𝑛1 = 𝑛2 = for one-sided
𝑝1 − 𝑝2 2

n1 = minimum sample size needed for group with predictor (+)


n2 = minimum sample size needed for group with predictor (-)
Total sample size needed = n1 + n2
Z1 – /2 = Z score for 1 – /2 (two-sided)
Z1 –  = Z score for 1 –  (one-sided)
Z1 –  = Z score for 1 – 
p1 = proportion of outcome (+) among group with predictor (+)
p2 = proportion of outcome (+) among group with predictor (-)
p1 – p2 = difference of proportion considered by the researcher to be clinically significant
p = p1+p2/2
q =1–p

Make sure n1p1q1 ≥ 5 and n2p2q2 ≥ 5

Alternative method:
Suppose we do not determine p1 – p2 (directly), instead we will use relative risk (RR) from
previous study to determine p1 – p2 (indirectly). After determining the p2 from previous
study or a pilot study, we can calculate p1, thus we will have result for p1 – p2
𝑝1
𝑅𝑅 =
𝑝2
Analytical study of an unpaired-dichotomous variable (case control study)

Outcome/disease (+) Outcome/disease (-) Total


Predictor/exposure (+) a b a+b
Predictor/exposure (-) c d c+d
Total a+c b+d a+b+c+d

2
𝑍1−𝛼 2 2𝑝𝑞 +𝑍1−𝛽 𝑝 1 𝑞 1 +𝑝 2 𝑞 2
𝑛1 = 𝑛2 = for two-sided
𝑝1 − 𝑝2 2

2
𝑍1−𝛼 2𝑝𝑞 +𝑍1−𝛽 𝑝 1 𝑞 1 +𝑝 2 𝑞 2
𝑛1 = 𝑛2 = for one-sided
𝑝1 − 𝑝2 2

n1 = minimum sample size needed for case/group with outcome (+)


n2 = minimum sample size needed for control/group with outcome (-)
Total sample size needed = n1 + n2
Z1 – /2 = Z score for 1 – /2 (two-sided)
Z1 –  = Z score for 1 –  (one-sided)
Z1 –  = Z score for 1 – 
p1 = proportion of predictor (+) among group with outcome (+)
p2 = proportion of predictor (+) among group with outcome (-)
p1 – p2 = difference of proportion considered by the researcher to be clinically significant
p = p1+p2/2
q =1–p

Make sure n1p1q1  5 and n2p2q2  5

Alternative method:
Suppose we do not determine p1 – p2 (directly), instead we will use odds ratio (OR) from
previous study to determine p1 – p2 (indirectly). After determining the p2 from previous
study or a pilot study, we can calculate p1, thus we will have result for p1 – p2

𝑝1 𝑞1 𝑝1 𝑞2 𝑝1 1 − 𝑝2
𝑂𝑅 = = =
𝑝2 𝑞2 𝑝2 𝑞1 𝑝2 1 − 𝑝1

Analytical study of an unpaired-continuous variable

2
𝑍1−𝛼 2 +𝑍1−𝛽 𝑠
𝑛1 = 𝑛2 = 2 × for two-sided
𝑥 1 −𝑥 2

2
𝑍1−𝛼 +𝑍1−𝛽 𝑠
𝑛1 = 𝑛2 = 2 × for one-sided
𝑥 1 −𝑥 2

n1 = minimum sample size needed for group 1


n2 = minimum sample size needed for group 2
Total sample size needed = n1 + n2
Z1 – /2 = Z score for 1 – /2 (two-sided)
Z1 –  = Z score for 1 –  (one-sided)
Z1 –  = Z score for 1 – 
x1 – x2 = difference of mean considered by the researcher to be clinically significant
s = combined standard deviation from previous study

if s were determined from a pilot study, then:


𝑠1 2 𝑛1 − 1 + 𝑠2 2 𝑛2 − 1
𝑠=
𝑛1 + 𝑛2 − 2
Analytical study of a paired-continuous variable
2
𝑍1−𝛼 2 +𝑍1−𝛽 𝑠
𝑛1 = 𝑛2 = for two-sided
𝑥 1 −𝑥 2

2
𝑍1−𝛼 +𝑍1−𝛽 𝑠
𝑛1 = 𝑛2 = for one-sided
𝑥 1 −𝑥 2

n1 = minimum sample size needed for group 1


n2 = minimum sample size needed for group 2
Total sample size needed = n1 + n2
Z1 – /2 = Z score for 1 – /2 (two-sided)
Z1 –  = Z score for 1 –  (one-sided)
Z1 –  = Z score for 1 – 
x1 – x2 = difference of mean considered by the researcher to be clinically significant
s = combined standard deviation from previous study

if s were determined from a pilot study, then:


𝑠1 2 𝑛1 − 1 + 𝑠2 2 𝑛2 − 1
𝑠=
𝑛1 + 𝑛2 − 2

Correlation study between two continuous variables

2
𝑍1−𝛼 2 +𝑍1−𝛽
𝑛= 1+𝑟 +3 for two-sided
0.5 𝑙𝑛
1−𝑟

2
𝑍1−𝛼 +𝑍1−𝛽
𝑛= 1+𝑟 +3 for one-sided
0.5 𝑙𝑛
1−𝑟

n = minimum sample size needed


Z1 – /2 = Z score for 1 – /2 (two-sided)
Z1 –  = Z score for 1 –  (one-sided)
Z1 –  = Z score for 1 – 
r = Pearson correlation coefficient from previous study or a pilot study

Notes
1. Do not forget to consider the non-response or dropout!
2. For descriptive study of a dichotomous variable, if the researcher were unable to find p
(proportion) from previous study or performing a pilot study, she could assign p = 0.5,
thus it would yield maximum result for pq. Remember to make sure npq  5 by
choosing appropriate d (precision).
3. For descriptive study of a continuous variable, if the researcher were unable to find s
(standard deviation) from previous study or performing a pilot study, an alternative
way to get it is by finding quarter of range result for outcome variable in population of
interest. For instance, the researcher wanted to know the mean Hb level (gr/dl) for
adult person in Bandung. The normal range of Hb level for adult in Hasan Sadikin
Hospital was 12 – 18 gr/dl. Roughly, she could estimate s (standard deviation) as (18 –
12)/4 = 1.5 gr/dl.
4. To estimate sample size with outcome variable numeric in order (APGAR score,
depression score, GCS, etc), they can be treated as continuous variable.
5. To estimate sample size with outcome variable ordinal (mild, moderate, severe), they
can be treated as dichotomous variable (mild vs moderate-severe or mild-moderate vs
severe)

You might also like