You are on page 1of 40

Sample Size

Calculation
Sample Size

Learning Objectives
 Importance of sample size selection
 Sampling risks ( risk &  risk) and drift
sensitivity
 Derivation of sample size formula for Z
distribution
 Calculating sample size using Minitab
 Determining sample size using OC curves

Seagate Confidential 2 Supplier Six Sigma Modular Training


Sample Size

Importance of Sample Size Selection

 There is a minimum amount of information that


must be supplied to estimate a sample size.

 The larger the sample the more sensitive the


test.

Seagate Confidential 3 Supplier Six Sigma Modular Training


Sample Size

The Effect of Sample Size


Standard Deviation of Sampled Mean
n=20


n n=10

n=5

Distributions of x for different n

Seagate Confidential 4 Supplier Six Sigma Modular Training


Sample Size

The Effect of Sample Size


n=20 Using a sample
size of n=5, the
sample mean is
n=5 not very likely to
show a process
change.

Current Process
Process Change

Seagate Confidential 5 Supplier Six Sigma Modular Training


Sample Size

Sampling Risks
 is the risk of concluding that there is a
difference, when there really isn’t one.
Process Shift 


 
observed value

 1- is the Confidence Level for H0.

Seagate Confidential 6 Supplier Six Sigma Modular Training


Sample Size

Sampling Risks
 is the risk of failing to detect a difference,
when there really is one.
Process Shift 


 
observed value

Power of a test (1-) is the chance of


detecting a difference, when there really is
one.

Seagate Confidential 7 Supplier Six Sigma Modular Training


Sample Size

Sampling Risks
 H0 : There is no difference
 H1 : There is significant difference

Null Hypothesis
Decision True False
Correct Type II Error
Accept H0 Decision

1-

Type I Error Correct


Reject H0 Decision

1-

Seagate Confidential 8 Supplier Six Sigma Modular Training


Sample Size

Drift Sensitivity
 Before conducting the experiment, you
need to pre-designate what may be
deemed to be a significant difference .

 This difference is then scaled by the


current standard deviation  to form a
dimensionless measure called the drift
sensitivity  /.
Which ?

Seagate Confidential 9 Supplier Six Sigma Modular Training


Sample Size

Making the Wrong Decision:  and  Error


 Lets say we are only willing to accept a =5% chance
of saying there is a difference when none really exist.
(Type I error)
 Furthermore, we are only willing to accept a =10%
chance of failing to detect a difference when one really
exists. (Type II error)

= 10%
/2 = 2.5%

X
Ho:  = o Ho:   o

Seagate Confidential 10 Supplier Six Sigma Modular Training


Sample Size

Sample Size Formula for Z Distribution


= 10%
/2 = 2.5%
Split the two
distributions
0 Ho:  = o Ho:   o x

 o  Z0.025  n k
o
Z0.025 x
n
k
 o    Z0.10  n k

Z 0 .1  x
Seagate Confidential n 11 Supplier Six Sigma Modular Training
Sample Size

Sample Size Formula for Z Distribution


 Z   Z 
o 0.025
x
0.1
x

n o
n

Z 0.025  x
 Z 0.1  x

n n

Z 0.025  Z 0.1  x  
n

Z 0.025  Z 0.1  x  n


2
 
n  Z 0.025  Z 0.1   x 
Z0.025  1.96
2
  
 
Z0.10  1.28
n
Z 0.025  Z 0.1 
2

k2
Seagate Confidential 12 Supplier Six Sigma Modular Training
Sample Size

Necessary Information
The difference, , you want to detect

 The  and  levels, typically 5% and
10%
 A GUESSTIMATE of the amount of
variability involved


n
Z0.025  Z0.1  x 2   ?
2 Most overlooked
piece of information
2  ?
ˆ  ?
Seagate Confidential 13 Supplier Six Sigma Modular Training
Sample Size

Exercise :(Risks, Drift and Sample Size)


Case    n


Constant Constant
?


Constant
? Constant


Constant ? Constant

? Constant Constant
4
Constant ? Constant
5
? Constant Constant
6
Seagate Confidential 14 Supplier Six Sigma Modular Training
Sample Size

Calculating Sample Size using


Minitab

Seagate Confidential 15 Supplier Six Sigma Modular Training


Sample Size

Tests for Comparison of Means


1-Sample Tests
 Stat  Power and Sample Size  1-Sample Z
 Stat  Power and Sample Size  1-Sample t

2-Sample Tests
 Stat  Power and Sample Size  2-Sample t

Seagate Confidential 16 Supplier Six Sigma Modular Training


Sample Size

Example 1 (1-Sample Tests)


The volume for the packaging of ice-cream cannot
vary more than 3 oz for a 64 oz container. The
packaging machines have a process variation of 1
oz.

What is the sample size required to estimate the


mean package volume at a confidence level of 99%
for power values of 0.7, 0.8 and 0.9?

Seagate Confidential 17 Supplier Six Sigma Modular Training


Sample Size

Example 1 (1-Sample Z-Test)


Stat  Power and Sample Size  1-Sample Z

Seagate Confidential 18 Supplier Six Sigma Modular Training


Sample Size

Example 1 (1-Sample t-Test)


Stat  Power and Sample Size  1-Sample t

Seagate Confidential 19 Supplier Six Sigma Modular Training


Sample Size

Example 1 (1-Sample Z vs 1-Sample t)


Power and Sample Size
Testing mean = null (versus not = null)
Calculating power for mean = null + 3 1-
Alpha = 0.01 Sigma = 1

1-Sample Z Test 1-Sample t Test


Target Actual Sample Actual Sample
Power Power Size Power Size
0.7000 0.9522 2 0.895 5
0.8000 0.9522 2 0.895 5
0.9000 0.9522 2 0.983 6

Seagate Confidential 20 Supplier Six Sigma Modular Training


Sample Size

Example 2 (2-Sample Tests)


An engineer plans to compare the effectiveness of a
new machine. The response of interest for the current
machine has a mean and standard deviation of 5 units
and 1 unit respectively.

He will recommend purchase of this new machine if its


mean response is at least 3 units higher than the
existing machine. Due to the cost of investment, he
wants to limit his risk of a bad investment to 5%.

What is the required sample size for power values of


0.6, 0.7 and 0.8?

Seagate Confidential 21 Supplier Six Sigma Modular Training


Sample Size

Example 2 (2-Sample t-Test)


Stat  Power and Sample Size  2-Sample t

Seagate Confidential 22 Supplier Six Sigma Modular Training


Sample Size

Tests for Comparison of


Proportions
1 Proportion Test
 Stat  Power and Sample Size  1 Proportion

2 Proportions Test
 Stat  Power and Sample Size  2 Proportions

Seagate Confidential 23 Supplier Six Sigma Modular Training


Sample Size

Example 3 (1 Proportion Test)


An executive wants to assess the Six Sigma
awareness at the local site. He wants to verify if at
least 50% of his staff is familiar with the Six Sigma
philosophy – an awareness program will have to be
launched if the actual awareness is less than 45%.

What sample size is require for =0.05 and =0.15?

Seagate Confidential 24 Supplier Six Sigma Modular Training


Sample Size

Example 3 (1 Proportion Test)


Stat  Power and Sample Size  1 Proportion

Seagate Confidential 25 Supplier Six Sigma Modular Training


Sample Size

Example 3 (1 Proportion Test)


Power and Sample Size
Test for One Proportion
Testing proportion = 0.5 (versus < 0.5)
Alpha = 0.05
1-
Alternative Sample Target Actual
Proportion Size Power Power
0.450000 717 0.8500 0.8504

Seagate Confidential 26 Supplier Six Sigma Modular Training


Sample Size

Example 4 (2 Proportions Test)


The executive wants to assess the effectiveness of the
new Six Sigma awareness program conducted.

Pre-program awareness level was estimated at 40%.


The program is deemed successful if the Six Sigma
awareness level is raised by at least 30%.

Determine the sample size required for =0.05 and


=0.15.

Seagate Confidential 27 Supplier Six Sigma Modular Training


Sample Size

Example 4 (2 Proportions Test)


Stat  Power and Sample Size  2 Proportions

Seagate Confidential 28 Supplier Six Sigma Modular Training


Sample Size

Example 4 (2 Proportions Test)


Power and Sample Size
Test for Two Proportions
Testing proportion 1 = proportion 2 (versus >)
Calculating power for proportion 2 = 0.4
Alpha = 0.05 1-
Sample Target Actual
Proportion 1 Size Power Power
0.700000 39 0.8500 0.8572

Seagate Confidential 29 Supplier Six Sigma Modular Training


Sample Size

Example 5 (One Way ANOVA)

Suppose you are about to undertake an investigation


to determine whether or not 4 treatments affect the
yield of a product. You would like to find significant
differences of +4. Thus, the maximum difference you
are considering is 4 units. Previous research suggests
the population  is 1.64.
Determine the sample size required for each treatment
using power of 0.9.

Seagate Confidential 30 Supplier Six Sigma Modular Training


Sample Size

Example 5 (One Way ANOVA)


Stat  Power and Sample Size One-Way ANOVA

Seagate Confidential 31 Supplier Six Sigma Modular Training


Sample Size

Example 5 (One Way ANOVA)

Power and Sample Size


One-way ANOVA
Sigma = 1.64 Alpha = 0.05 Number of Levels = 4
Sample Target Actual Maximum
SS Means Size Power Power Difference
8 6 0.9000 0.9096 4

Seagate Confidential 32 Supplier Six Sigma Modular Training


Sample Size

Calculating Sample Size using


Operating Characteristics
Curves

 F – Test and 2 Test

Seagate Confidential 33 Supplier Six Sigma Modular Training


Sample Size

OC Curves for 2-sided F-Test


Probability of accepting Ho

n1 = n
n21= =
3 n2 =
3


OC curves for different values of n for the 2-sided F-test for a
level of significance =0.05
Seagate Confidential 34 Supplier Six Sigma Modular Training
Sample Size

Probability of accepting Ho OC Curves for 1-sided F-Test


n1 = n2
= 2


OC curves for different values of n for the 1-sided F-test for a
level of significance =0.05
Seagate Confidential 35 Supplier Six Sigma Modular Training
Sample Size

OC Curves for 2-sided


Probability of accepting Ho
 2
Test
= 2
n

n=
2


OC curves for different values of n for the 2-sided  2
test for a
level of significance =0.05
Seagate Confidential 36 Supplier Six Sigma Modular Training
Sample Size

 2
OC Curves for 1-sided (Upper Tail) Test
Probability of accepting Ho

n=
2
n=2


OC curves for different values of n for the 1-sided (upper tail) 2
test for a level of significance =0.05
Seagate Confidential 37 Supplier Six Sigma Modular Training
Sample Size

 2
OC Curves for 1-sided (Lower Tail) Test
Probability of accepting Ho
2
=
n


OC curves for different values of n for the 1-sided (lower tail) 2
test for a level of significance =0.05
Seagate Confidential 38 Supplier Six Sigma Modular Training
Sample Size

OC Curves for Various Tests

The OC Curves for various tests including the 5 curves


shown in the powerpoint slides can be found in the
following book:

Title: Applied Statistics and Probability for Engineers


Author: Douglas C Montgomery, George C Runger
Publisher: John Wiley & Sons, Inc (2 nd Edition)
Page: Appendix A12 – A28

Seagate Confidential 39 Supplier Six Sigma Modular Training


Sample Size

End of Topic
What question do you have?

Seagate Confidential 40 Supplier Six Sigma Modular Training

You might also like