Lean Six Sigma Operations

Black Belt – Week 1

Sample Size Calculation

Release 1.0
Objectives
• Introduce the topic of sample size determination as a
way to manage risk
– Alpha
– Beta
– Delta
– Sample Size
• Demonstrate how to calculate sample sizes in
Minitab
– 1-sample t
– One-way ANOVA
– 2-sample proportions

For Ingersoll Rand Internal Use Only

2 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Lean Sigma Roadmap - Measure
Steps Tools Outputs
• •SIPOC
SIPOC/ /VSM
VSM
• •Input/Output
Input/Output
Analysis
Understand the
Analysis  Current State
• •C&E
Process C&EMatrix
Matrix Process Maps
• •Detailed
DetailedProcess
Process
Maps
Maps  Identified and
measured X’s
Measure • •Data
DataCollection
CollectionPlans
Plans (KPIV’s)
• •Data
• Understand the DataIntegrity
IntegrityAudits
Audits  Measurement
Process Develop and Evaluate • •Continuous
ContinuousMSA
MSA
• Develop and
Evaluate
Measurement Systems (Gage
(GageR&R)
R&R)
system
Measurement
Systems
• •Attribute
AttributeMSA
MSA(Kappa
(Kappa verified
Studies)
Studies)
• Measure Current
Process
 Current
Performance capability of
Measure Current
• •Process
Y’s (KPOV’s)
Process Performance ProcessCapability
Capability

For Ingersoll Rand Internal Use Only

3 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Why Sample?
It is not always feasible or possible to analyze 100% of the population...

But, we can draw conclusions about the population

through statistically relevant samples
For Ingersoll Rand Internal Use Only
4 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
What is a Sample?
• Collecting only a portion of the data that is available
or could be available
• Using the data from the sample to draw statistical
inferences
• A faster, less costly way to gain insight into a process
or large population
• Remember: Even if we collect 100% of the data
points for our process or population during a certain
period of time, it is still only a sample of the whole
population

AAgood
goodsample
sampleis isaaminiature
miniatureversion
versionof
ofthe
the
population
population--just
justlike
likeit,
it,only
onlysmaller.
smaller.

For Ingersoll Rand Internal Use Only

5 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
When Does Sampling Work?
• Each member of the population has an equal chance
of being selected
• Selecting one member doesn’t influence likelihood of
another member being selected or not
• There aren’t any significant differences between
those selected and those that weren’t
• You have a large enough sample to find what you’re
looking for. If it’s rare or you want to be very precise,
you’ll need a large sample

For Ingersoll Rand Internal Use Only

6 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Process or Population?
What is a Process and what is a Population

A Process . . . A Population . . .
• Back guard assembly • All the back guards produced
• Product made using one
• Porcelain spraying supplier lot
• Deals that exceed margin
• Quoting activity targets
• Items with discrepancies
• Inventory counts maintenance

Process
Processvs.
[Link]
Populationdetermines
determinessampling
samplingapproach,
approach,
sample
samplesize
sizecalculations,
calculations,and
andother
otherconsiderations.
considerations.
For Ingersoll Rand Internal Use Only
7 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Sampling:
How, What, When

Purpose Considerations Sample Size Approach

Process  Take action or  Where to sample  Use  Subgroup Sampling
predict future guidelines to
 Frequency  Systematic
- In Control? analysis tool
Sampling
 Grouping selection
- Capable?
 Representative
- Improve?
 Cost

Large  Describe or  Precision (+/- )  Use Minitab  Random Sampling

Population Quantify procedures
 Amount of  Stratified Random
Characteristics
characterisitic’s Sampling
variation:  or P
 Systematic
- Confidence level Sampling
- Representative  Cluster Sampling
- Cost

Reference
ReferenceMatrix
Matrix
For Ingersoll Rand Internal Use Only
8 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
Introduction - The Learning Process
“An initial hypothesis leads by a process of deduction to certain
necessary consequences that may be compared with data.

When consequences and data fail to agree, the discrepancy can

lead, by a process called induction, to a modification of the
hypothesis.

A second cycle in the investigation is thus initiated.”

– Box, Hunter & Hunter

A hypothesis is the translation of a practical question

into a statistical question

For Ingersoll Rand Internal Use Only

9 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
Review of Hypothesis Testing
• A hypothesis is a statement of what we believe might be
true.
• In general, we formulate two competing hypotheses:
– A null hypothesis (Ho) is a statement of no change.
– An alternate hypothesis (Ha) is a statement that a change (or
difference) exists.
• We conduct an experiment that seeks to reject the null
hypothesis.
• By analyzing the experimental data using statistical tools
(e.g., DOE, ANOVA, t-test), we decide whether to reject
the null hypothesis.
• If the null hypothesis is rejected, then we accept the
alternative hypothesis as true.

For Ingersoll Rand Internal Use Only

10 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
Hypotheses and Decision Risk
• There are two kinds of decision risk in hypothesis
testing:
– Rejection of the null hypothesis when it is true
(Type I error).
– Acceptance of the null hypothesis when it is false
(Type II error).
• We control the decision risk by specifying, in advance
of the experiment, how much risk we can tolerate.
• After specifying our tolerance for risk, we have the
information needed to determine the optimal sample
size.

For Ingersoll Rand Internal Use Only

11 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
The Sample Size Cookbook
• Define problem
• Develop objectives
• Establish hypotheses
• Design test
 Establish alpha
d ay ’s Establish beta
To 

Establish delta
T op i c s 

 Establish sample size

 Devise sampling plan
• Select sample
• Conduct test
• Measure and record data
• Conduct statistical test
• Make statistical decision
• Translate decisions to action

For Ingersoll Rand Internal Use Only

12 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Risk
• Our choice of a determines when to reject the null hypothesis:
– If P>, fail to reject null hypothesis (no change)
– If P< , reject null hypothesis (there is a change)
• There is a risk of rejecting the null hypothesis when it, in fact, is
true (Type I error). The probability that this happens is exactly
equal to .
• A Type I error means that the action we take will, unfortunately,
generate no improvement
• The  value to use should depend upon practical
considerations-- financial risk, safety risk, or other risk to the
customer
• Typical values are  = 0.05 - 0.10

For Ingersoll Rand Internal Use Only

13 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Risk
•  is the risk of accepting the null hypothesis when it is actually
false (Type II error).
• A Type II error means that we overlook an opportunity for
improvement - we take no action when we should.
• This risk can be the greatest deterent to breakthrough and must
be seriously considered.
• Typical values in industrial experiments are usually  = 0.10 -
0.20
• The power of a test is equal to 1-. High powers (say 0.8 or
larger) give a good chance that we will not overlook
opportunities for improvement

For Ingersoll Rand Internal Use Only

14 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
Evaluation of Decision Error

Your
YourDecision
Decision

Accept Ho Reject Ho

Type I
Ho True Correct
Correct Error

The -Risk)
The
Truth
Truth Type II
Error Correct
Correct
Ho False
-Risk)

For Ingersoll Rand Internal Use Only

15 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
Signal-to-Noise Ratio (/ )
•  is the size of the minimum change (effect) that we want to be
reasonably sure of detecting during the experiment ( is also
known as the signal).
 is the standard deviation of the process (it is also known as
the noise).
• The signal-to-noise ratio, / , is directly related to our ability to
detect a process change.
• High signal-to-noise ratios are associated with good statistical
power (ability to detect a change when such a change occurs).

For Ingersoll Rand Internal Use Only

16 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Simple Random Sample
• If all possible samples of n experimental units are equally likely, then
the procedure is a simple random sample.
• Example: We generate 15 random numbers from 1 to the number of
items produced in a shift for the enamel baking process
– These items identified are then evaluated with respect to the output
measures
• Characteristics of SRS
– Unbiased: Every experimental unit has the same chance of being
chosen
– Independence: The selection of an experimental unit is not dependent
on the selection of another
• In continuous processes, using a SRS is very difficult because there
is no clear experimental unit. Generally we sample from a stream of
material.

For Ingersoll Rand Internal Use Only

17 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Stratified Sample
• Divide the population into homogeneous groups and randomly
sample from within each group
• Example: Stamping Batches
– There are a large number of batches (batched by steel lot) going
through the stamping press with each press batch having many
items created
– Randomly sample 2 items from each batch
– This sample will effectively represent each batch’s effect on the
variability of the output variables
• Another example
– If you have 2 paint lines, randomly sample product from each line
– This sample allows you to represent the effect of each line on the
variability of the output

For Ingersoll Rand Internal Use Only

18 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Cluster Sample

• Divides the sample into smaller homogeneous groups,

then the groups are randomly sampled
• Example: Back to Paint
– Assume there are a number of positions for product within
each paint line
– Number each paint line and randomly sample a subset of
the paint lines
– Then randomly sample positions within the selected paint
lines
• This represents line effects without having to sample from
all lines

For Ingersoll Rand Internal Use Only

19 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Systematic Sample

• Start with a randomly chosen unit and sample every kth

unit thereafter
• Example:
– For a assembly shift, there are N number of product
produced
– Pick a number at random to start and then select every 15th
unit to evaluate
• This method is good because its simple

For Ingersoll Rand Internal Use Only

20 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Subgroup Sample
• Sample output of step or activity with some frequency -
usually a time increment
• Remember Sources of Variation -
– Will subgroup show variation you’re interested in?
– Frequency - predetermined, corresponds to control chart
sampling
• Example:
– Pull 5 ice machines at 10 a.m., 12 p.m., 2 p.m., and 4 p.m.

For Ingersoll Rand Internal Use Only

21 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
Standard Sampling Strategies
Which Strategy Is Right For My Project?
• Consider what main factors or Sources of Variation are on your
suspect list
– Will your strategy allow you to see what you want to see?
• The more complex the strategy, the harder it is to plan, execute,
and control
– Only use the degree of sophistication you really need
• Sampling can be a complicated question
– Review your approach with a mentor or Master before you
start data collection

When In Doubt, Ask For Help!

For Ingersoll Rand Internal Use Only

22 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Sample Size (N)
• N is the sample size that we will collect at each level of an
experimental factor

• Example: In any 2k experiment, we have 2 levels per factor, so:

– The total number of responses that are observed in the experiment

is 2*N

– The number of times each experimental combination is replicated

is (2*N)/2k

For Ingersoll Rand Internal Use Only

23 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Calculating Sample Size
• The relationship among the five variables and
N can be expressed by an equation like this:

(Z  / 2  Z ) 2

N
  2

• Note that if we know any four of the five variables, we

can calculate the fifth
• Using Minitab, we will now show how to calculate
sample sizes and powers:
– Sample Size = N
– Statistical Power = 1-

For Ingersoll Rand Internal Use Only

24 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
  20%  10%  5%  1%
 20% 10% 5% 1% 20% 10% 5% 1% 20% 10% 5% 1% 20% 10% 5% 1%

0.2 225 328 428 651 309 428 541 789 392 525 650 919 584 744 891 1202
0.3 100 146 190 289 137 190 241 350 174 234 289 408 260 331 396 534
0.4 56 82 107 163 77 107 135 197 98 131 162 230 146 186 223 300
0.5 36 53 69 104 49 69 87 126 63 84 104 147 93 119 143 192
0.6 25 36 48 72 34 48 60 88 44 58 72 102 65 83 99 134
0.7 18 27 35 53 25 35 44 64 32 43 53 75 48 61 73 98 Old-Fashioned Way
0.8 14 21 27 41 19 27 34 49 25 33 41 57 36 46 56 75
0.9 11 16 21 32 15 21 27 39 19 26 32 45 29 37 44 59 of Getting Sample
1.0
1.1
9
7
13
11
17
14
26
22
12
10
17
14
22
18
32
26
16
13
21
17
26
21
37
30
23
19
30
25
36
29
48
40 Size
1.2 6 9 12 18 9 12 15 22 11 15 18 26 16 21 25 33
1.3 5 8 10 15 7 10 13 19 9 12 15 22 14 18 21 28
1.4
1.5
5
4
7
6
9
8
13
12
6
5
9
8
11
10
16
14
8
7
11
9
13
12
19
16
12
10
15
13
18
16
25
21
(Before the advent of
1.6
1.7
4
3
5
5
7
6
10
9
5
4
7
6
8
7
12
11
6
5
8
7
10
9
14
13
9
8
12
10
14
12
19
17
Minitab)
1.8 3 4 5 8 4 5 7 10 5 6 8 11 7 9 11 15
1.9 2 4 5 7 3 5 6 9 4 6 7 10 6 8 10 13
2.0 2 3 4 7 3 4 5 8 4 5 6 9 6 7 9 12
2.1 2 3 4 6 3 4 5 7 4 5 6 8 5 7 8 11
2.2 2 3 4 5 3 4 4 7 3 4 5 8 5 6 7 10
2.3 2 2 3 5 2 3 4 6 3 4 5 7 4 6 7 9
2.4 2 2 3 5 2 3 4 5 3 4 5 6 4 5 6 8
2.5 1 2 3 4 2 3 3 5 3 3 4 6 4 5 6 8
2.6 1 2 3 4 2 3 3 5 2 3 4 5 3 4 5 7
2.7 1 2 2 4 2 2 3 4 2 3 4 5 3 4 5 7
2.8 1 2 2 3 2 2 3 4 2 3 3 5 3 4 5 6
2.9 1 2 2 3 1 2 3 4 2 2 3 4 3 4 4 6
3.0 1 1 2 3 1 2 2 4 2 2 3 4 3 3 4 5
3.1 1 1 2 3 1 2 2 3 2 2 3 4 2 3 4 5
3.2 1 1 2 3 1 2 2 3 2 2 3 4 2 3 3 5
3.3 1 1 2 2 1 2 2 3 1 2 2 3 2 3 3 4
3.4 1 1 1 2 1 1 2 3 1 2 2 3 2 3 3 4
3.5 1 1 1 2 1 1 2 3 1 2 2 3 2 2 3 4
3.6 1 1 1 2 1 1 2 2 1 2 2 3 2 2 3 4
3.7 1 1 1 2 1 1 2 2 1 2 2 3 2 2 3 4
3.8 1 1 1 2 1 1 1 2 1 1 2 3 2 2 2 3
3.9 1 1 1 2 1 1 1 2 1 1 2 2 2 2 2 3
4.0 1 1 1 2 1 1 1 2 1 1 2 2 1 2 2 3

For Ingersoll Rand Internal Use Only

25 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Example:
Welder, 1-Sample t-Test

• A new type of Welder has been developed

– To be economically worthwhile, the new Welder must have a yield that is
at least five (5) percentage points greater than the current model
– The current Welder has a yield of about 80%
• To compare the two Welders, daily data will be used
– The historic daily yield of the current Welder has a standard deviation of
three (3) percentage points
– The new Welder is expected to have the same daily yield variation
• The a risk that is acceptable to us (chance of saying the two Welders
are different when they really aren‘t) is 0.01
• The b risk that is acceptable to us (chance of saying the two Welders
are the same when they really aren‘t) is 0.20

For Ingersoll Rand Internal Use Only

26 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
Example:
Welder - Hypotheses
 NullHypothesis (Ho): The
average daily yield of the new
Welder is LESS THAN OR
EQUAL TO the average daily
yield of the current Welder

 AlternateHypothesis (Ha): The

average daily yield of the new
Welder is GREATER THAN the
average daily yield of the
current Welder

For Ingersoll Rand Internal Use Only

27 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
Example:
Welder - Sample Size

 What sample size is required to choose between the

two hypotheses?

 From the information given, we have

 Significance level,  = 0.01

 Power = 1 = 0.80
 Difference in means,  = 85  80 = 5
 Standard Deviation,  = 3

 Alternate Hypothesis (Ha) is that yield of the new

Welder is greater than yield of current Welder

For Ingersoll Rand Internal Use Only

28 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Example:
Welder - Using Minitab
Select Stat >Power and Sample Size > 1-Sample t. Click
third button “Calculate sample size for each power value.”

2. Enter delta,  5. Enter

alpha, 
1. Enter power, 

3. Enter sigma, 

4. Select alternate Fill in items 1- 3. Then click “Options”

hypothesis and fill in items 4 & 5. Finally, click OK
in all open windows.
For Ingersoll Rand Internal Use Only
29 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Example:
Welder - Minitab Output

Minitab informs us that we need a sample of size seven (7) from the
new Welder to ensure, with required power 0.80, that we accept it if
five percentage points better than the current Welder
(Note: N=7 actually gives a power of 0.8512)

For Ingersoll Rand Internal Use Only

30 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Example:
Reject Parts

• A company is trying to decide whether to switch suppliers

• Supplier 1, the current supplier, has a reject rate around
10%
• Supplier 2, a new supplier, promises to deliver parts that
have a reject rate less than 10%
• The customer believes it would be economical to switch if
the reject rate of Supplier 2 is at least one (1) percentage
point smaller than Supplier 1
• What is the size of the sample that should be taken to
determine whether it is worth switching to Supplier 2?
• For purposes of comparison, calculate the required
sample sizes to achieve statistical powers of 0.6, 0.7, 0.8,
and 0.9 (Assume a = 0.05)
For Ingersoll Rand Internal Use Only
31 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
Example:
Reject Parts - Hypotheses
 Null Hypothesis (Ho): The
proportion of reject parts from
Supplier 1 IS EQUAL TO
proportion of reject parts from
Supplier 2

 Alternate Hypothesis (Ha): The

proportion of reject parts from
Supplier 1 IS GREATER THAN the
proportion of reject parts from
Supplier 2

For Ingersoll Rand Internal Use Only

32 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
Example:
Reject Parts - Sample Size

 What sample sizes are needed at powers = 0.6, 0.7,

0.8, and 0.9 to compare two observed proportions?

 From the information given, we have

 Significance level,  = 0.05

 Customer will choose Supplier Y if the proportion
of bad parts is at least one (1) percentage point
below Supplier X
 Difference in Proportions,  = 0.01
 Alternate Hypothesis (Ha) is that the Supplier X
Reject Rate is greater than the Supplier Y Reject
Rate

For Ingersoll Rand Internal Use Only

33 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Example:
Reject Parts - Minitab Output
Select Stat >Power and Sample Size > 2 Proportions. Click
third button “Calculate sample size for each power value.”

2. Enter Supplier
X proportion
1. Enter power
values 5. Enter alpha

3. Enter Supplier
Y proportion
4. Specify Supplier X Fill in items 1-3. Then click
greater than Supplier Y “Options” and fill in items
4&5. Finally, click OK in all
open windows.
For Ingersoll Rand Internal Use Only
34 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Example:
Reject Parts - Minitab Output
Power and Sample Size

Test for Two Proportions

Testing proportion 1 = proportion 2 (versus >)

Calculating power for proportion 2 = 0.09
Alpha = 0.05
Sample Target
Proportion 1 Size Power Actual Power
0.1 6196 0.6 0.600035
0.1 8091 0.7 0.700008
0.1 10630 0.8 0.800005
0.1 14724 0.9 0.900006

Large sample sizes are required! To give a 90% confidence in

detecting a difference of one percentage point (if it exists), samples
of 14724 parts are required from each supplier!
For Ingersoll Rand Internal Use Only
35 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 More Fun with Minitab
• Draw the arrow in Ca se    n
the up or down
?

Constant Constant
direction according
to each scenario
• For example, in 
Constant ? Constant

Case 1, if  and 
are constant and ?

Constant Constant
you’re interested in
a smaller , will
Constant Constant
your sample size 4
get larger (up
arrow) or smaller Constant ? Constant
(down arrow)?
5
• Use Minitab to play ? Constant Constant
6
around with the
numbers
For Ingersoll Rand Internal Use Only
36 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 BB Hints:
Sample Size By Tool
Tip Sheet
Tool or Statistic Minimum Sample Size
Average 5 - 10
Standard Deviation 25 - 30
Proportion Defective (P) 100 and nP >= 5
Histogram or Pareto 50
Scatter Diagram 25
Control Chart 20
Remember
Rememberthese
theseare
areMINIMUMS.
MINIMUMS. More
Moredata
data
points means higher confidence in
points means higher confidence in
conclusions
conclusionsyou
youcan
candraw
drawfrom
fromyour
yourdata
data
For Ingersoll Rand Internal Use Only
37 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
 Sampling Processes
Considerations To Keep In Mind
• Where you Sample
– Location in the process where your process variable affects the output variable
– Sample as far upstream as practically and logically possible
• Frequency of Sampling
– Often enough to catch it going from good to bad
– It is better to have several small samples over time than one large sample at
single point in time
– Sample more frequently if process stability is questionable
• Sub-grouping
– Minimize opportunity for special cause in subgroup
– Keep track of what, where, when, who, etc.
– For subgroups or Sources of Variation, you’ll need at least 5 data points for each
category of the variable

For Ingersoll Rand Internal Use Only

38 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
Common Issues
With Sample Quality
• Selection Bias
– No structure to determine which items you sample
– Convenience sampling; Systematic sampling matches some structure
• Changes in the Environment
– Environmental changes make the sample no longer representative
• Non-Response Bias
– Especially in surveys, characteristics of those who don’t respond are
different in a significant way from those who do (vocal minority)
• Measurement Bias
– Often related to unequal batch sizes - tendency to select larger batches to
sample from
– Non-representative batches . . . long overdue bills without sampling quickly
paid accounts
• Sampling Plan Executed Improperly
– Try to be aware of biases and other quality problems that can creep in and
affect your data!

For Ingersoll Rand Internal Use Only

39 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
Summary
• Introduced the topic of sample size determination as a way to
manage risk
– Alpha
– Beta
– Delta
– Sample Size
• Demonstrated how to calculate sample sizes in Minitab
– 1-sample t
– One-way ANOVA
– 2-sample proportions

For Ingersoll Rand Internal Use Only

41 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.
Connecting the Dots
• Statistics rely on the Central Limit Theorem to allow
us to make statistical inferences.
• Sample size calculations are required throughout the
Measure and Analyze phases to insure statistical
relevance and confirm that our inferences are valid.

For Ingersoll Rand Internal Use Only

42 Rev. B Printed 3/23/2009
© 2001 by Sigma Breakthrough Technologies, Inc.