Professional Documents
Culture Documents
• Types of Data
• Measures of the Center of the Data
– Mean
– Median
• Measures of the Spread of Data
– Range
– Variance
– Standard Deviation
• Properties of a Normal Distribution
10 n-1
N
Mean
square (X X )
i=1
i
2
n-1
Problem: Using the form above, calculate the standard
deviation for the numbers: 21354
8 Confidential, proprietary information of Tyco International Ltd.
Example
X X- X (X - X)2
1 2 -1 1
2 1 -2 4
3 3 0 0
4 5 2 4
5 4 1 1
6 N
7 (X X )
i=1
i
2
8 n-1
9
10
N
Mean
15
3
10
i
(X
i=1
X ) 2
Distribution
Two
Distribution Three
68% probability of
40%
obtaining a value
between two
30%
95% values
20%
99.73%
10%
0%
-4 -3 -2 -1 0 1 2 3 4
Number of standard deviations from the mean
13 Confidential, proprietary information of Tyco International Ltd.
Descriptive Statistics
• Let’s look at the different data sets
• We will start with some Descriptive Statistics
• Open Normal.mpj
Median
Median
60
50
40
30
20
10
5
0.1
1 2 3 4 5 6 7 8
Dist1
60
50
40
30
20
10
5
0.1
-5 0 5 10 15 20 25
Dist2
15
Frequency
10
0
45.0 52.5 60.0 67.5 75.0 82.5 90.0
Output
Mean
X = 266
There is variation from sample to sample . . .
So how can I use this data to reflect my population
(Standard Deviation)
C.I. = Statistic K *
n
Statistic = Mean, Variance, etc.
K = Constant based on the reference distributi on
t - Distribution
5 2.78 1.96
10 2.26 1.96
20 2.09 1.96
30 2.05 1.96
100 1.98 1.96
1000 1.96 1.96
-4 -3 -2 -1 0 1 2 3 4
s s
x t / 2,n1 x t / 2,n1
n n
becomes
s s
x 2.26 x 2.26
n n
33 Confidential, proprietary information of Tyco International Ltd.
CONFIDENCE INTERVAL FOR THE MEAN
• Let’s look at the Confidence Interval for
the mean
• Suppose we want to determine the 95%
CI for the mean from 10 samples from a
process. We sample the process and
get:
– Xbar = 249.6
– s = 14.15
– n = 10
• from the following data:
– 263.1, 249.2, 247.4, 263.7, 262.4, 255.6, 252.4,
251.5, 227.3, 223.0
14.15 14.15
249.6 2.26 249.6 2.26
10 10
239.5 259.7
We have 95% confidence that the true mean of the
population falls between 239.49 and 259.71
Confidence Intervals
Variable N Mean StDev SE Mean 95.0 % CI
Data 10 249.56 14.15 4.47 ( 239.44, 259.68)
16 1 16 1
1.66 1. 66
c(.2 05 / 2 ),15 c(21.05 / 2 ),15
15 15
1.66 1.66
c.2025,15 c.2975,15
1.66
15
27.49
1.66
15
6.26
1.23 2.57
37 Confidential, proprietary information of Tyco International Ltd.
What if my data is
Not normally distributed?
-Chi-Square
• Statistical tests are dependant upon
-Poisson the type of distribution that exists
-Binomial
-Hypergeometric • Using tools meant for normally-
-Weibull distributed data, when the distribution
is non-normal can lead to false
-Exponential
conclusions
-Uniform
-Lognormal • It is important to know, where
-Logistic possible, what type of distribution you
are dealing with.
2
Party Time
J F M A
1996 1997
Derived from Understanding Variation: The Key To Managing Chaos, Donald J. Wheeler, SPC Press. 1993.
Confidential, proprietary information of Tyco International Ltd.
Variation Example
Special vs. Common Cause
2
Manager wants to take back award
J F M A M J J
1996 1997
Derived from Understanding Variation: The Key To Managing Chaos, Donald J. Wheeler, SPC Press. 1993.
Confidential, proprietary information of Tyco International Ltd.
Variation Example
Special vs. Common Cause
• By November, scrap has risen to a value of 2.6% - Year High
• Manager decides to take action
• A “special meeting” is called to solve this problem once and for all
• After a sound lecture on the importance of scrap, the manager
leaves. Employees aren’t sure what to do. Besides, they have other
3 metrics which are more important. So they do nothing.
Scrap Level (%)
2
No more “Soft Management”
J F M A M J J A S O N D
1996 1997
Derived from Understanding Variation: The Key To Managing Chaos, Donald J. Wheeler, SPC Press. 1993.
Confidential, proprietary information of Tyco International Ltd.
Variation Example
Special vs. Common Cause
• Manager has seen reduced scrap levels since the end of last year.
“Things are looking-up!”
(REMEMBER: Nothing had been done to change the system)
• His take away: “A tough management style gets results!”
Manager concludes:
“Tough Love Makes Things Happen
Scrap Level (%)
J F M A M J J A S O N D J F M A M J
1996 1997
Derived from Understanding Variation: The Key To Managing Chaos, Donald J. Wheeler, SPC Press. 1993.
Confidential, proprietary information of Tyco International Ltd.
Variation Example
Special vs. Common Cause
3 UCL
1 LCL
J F M A M J J A S O N D J F M M J J A S O
1996 1997
Derived from Understanding Variation: The Key To Managing Chaos, Donald J. Wheeler, SPC Press. 1993.
52 Confidential, proprietary information of Tyco International Ltd.
The Control Chart Tells A Different Story
• Manager
“ Hey, I made my decision based on data - How can I be wrong ?”
• Black Belt
“Your decisions were made from observing high and low points as
signals, when in reality, it was all noise (Common Cause Variation).
Look at the data, there has been no significant change in the
process.”
Manager concludes, “Tough
Love Makes Things Happen!”
3 UCL
Scrap Level (%)
1 LCL
J F M A M J J A S O N D J F M M J J A S O
1996 1997
Party Time Manager Wants No more soft management
To Take Back Award
Common Cause
Special Cause
UCL
CHARACTERISTIC
MONITORED
Plotted Data
Center Line
LCL
UCL
LCL
• What % of data points should fall between the UCL and LCL ?
• If a point falls beyond the UCL or LCL does this mean we are
making a defect for the customer ?
% of Data Points
UCL
3 Sigma 99-99.9%
2 Sigma 90-98%
The Item
We Are 1 Sigma
Measuring 60-75%
1 Sigma
2 Sigma
3 Sigma
LCL
TIME
UCL
USL
LSL
LCL
TIME
UCL
LCL
LSL
TIME
Attribute Charts
p-Chart
np-Chart
c-Chart
u-Chart
ATTRIBUTE -
The data are generally counted
Results from using go/no-go gages,
inspection of visual defects, quantity of
missing parts, pass/fail or yes/no decisions,
etc.
Confidential, proprietary information of Tyco International Ltd.
Exercise: What Type of Data ?
Counting
Data Collected In
Specific Defects or
Groups or Individuals ?
Defective Items ?
GROUPS INDIVIDUAL
(Averages) VALUES Specific Defective
(n>1) (n=1) Types Of Items
“Defects”
X-Bar R Individuals
X-Bar S Moving Range Is The Probability Of If You Know How Many
A Defect Low ? Are Bad, Do You Know
NO
How Many Are Good?
NOTE: X-Bar S is appropriate
Poisson Distribution Binomial Distribution
for subgroup sizes (n) of > 10
YES YES
Individuals
Moving Range
Area of
Constant
Opportunity Constant
Sample Size ?
In Each Sample
Size ?
NO YES NO YES
U Chart C Chart P Chart NP Chart
STAT>CONTROL CHARTS>….
5 _
X=4.04
LC L=-2.05
2 4 6 8 10 12 14 16 18 20 22 24
O bser vation
8
U C L=7.488
6
M oving Range
4
__
2 M R=2.292
0 LC L=0
2 4 6 8 10 12 14 16 18 20 22 24
O bser vation