# Lies_ Damned Lies_ and Statistics by wanghonghx

VIEWS: 6 PAGES: 22

• pg 1
```									Lies, Damned Lies, and
Statistics
Hanford Performance Indicator Forum
December 11, 2003

Steven S Prevette
Fluor Hanford Safety and Health

509-373-9371
Steven_s_prevette@rl.gov
http://www.hanford.gov/safety/vpp/trend.htm
Introduction

• Numbers abound in the Safety profession
• We have all seen great uses and misuses

• How can we make the best use of the data
we have and make effective decisions that
will improve the safety of ourselves, our
coworkers, and our companies?
Liars Figure and Figures Lie 1

18
16                                                Injuries by Month
14
12
10
8
6
4
2
0
Jan-02

Jul-02

Jan-03

Jul-03

Jan-04
Apr-02

Oct-02

Apr-03

Oct-03
Here are the past 2 years of injury data.
What is your interpretation? What is your
prediction of future rates?
Liars Figure and Figures Lie 2

18
16                                                         Injuries by Month
14
12
10
8
6
4
2
y = -0.132x + 11.987
0
Jan-02

Jan-03

Jan-04
Jul-02

Jul-03
Apr-02

Oct-02

Apr-03

interpretation now? What is your
prediction of future rates?
Liars Figure and Figures Lie 3

18                                       Injuries by Month
16
14
12
10
8
6
4
2                 6 Month Moving Average
0
Jan-02

Jan-03

Jan-04
Jul-02

Jul-03
Apr-02

Oct-02

Apr-03

Oct-03
Let’s add a 6 month moving average. What
is your interpretation now? What is your
prediction of future rates?
Interlude – Fun with Charts

• It appears I can make this data say
anything I want it to say . . .
• Courses are taught on ―chartsmanship‖

• Can we interpret the data with a sound,
rigorous method that is repeatable and
gives credible results?
Liars Figure and Figures Lie 4
25                                                                   Average = 10.3
(Jan02 - Jan04)
20
Upper Control Limit
15

10

5
Lower Control Limit
0

Nov-02

Nov-03
Jan-02

Jan-03

Jan-04
Jul-02

Jul-03
May-02

Sep-02

May-03

Sep-03
Mar-02

How about the Control Chart?                                                       Mar-03

- The data are stable, within certain bounds.

• Many of you have experienced the ―Red
• Results changed from person to person,
but were random and predictable
• The data here were generated by the Red
• Which chart gave the proper interpretation
and predictions?
The Control Chart

• Separates random noise from signal
• Provides prediction capability
• When coupled with management theory
and action can lead to significant
improvements
• Does not require great costs to perform
• Actually has saved money
Management Theory For
Improvement

• When there is a trend, find ―special cause‖
and act upon the trend

• When there is no trend, find ―common
causes‖ and act upon the system, the
overall performance
Pareto Charts
OSHA Cases FY 1997
Fluor Hanford
OSHA Cases FY 1998 through May 1998
"Overall 25% Reduction to Date"
0.40

0.35

0.30

0.25

0.20

0.15     -48%                             -26%
-8%
72%
0.10
-63%
0.05

0.00

OR                  R        HP T              GE
R           IAN
R AT                TT E                        RIG            RIC
OPE             E FI                          E R/            ECT
S              PIP                            RK             EL
CES                                               WO
RO                                            IRO
N
ARP
CLE
NU

Pareto charts can be used to analyze
common causes – FH ―Top 5 Occupation‖
initiative from 1998
What is a Trend?

•   One point outside the control limits
•    Two out of Three points two standard deviations
above/below average
•    Four out of Five points one standard deviation
above/below average
•    Seven points in a row all above/below average
•    Ten out of Eleven points above/below average
•    Seven points all increasing/decreasing.

See http://www.hanford.gov/safety/vpp/trend.htm

•   This system can be used with leading indicators and
Behavior Based Safety

•    A personal example – my group targeted computer
ergonomics

•    Goal was to improve behaviors on the computer, to
reduce risk of Carpal Tunnel / Overuse Syndrome
BEHAVIOR TO OBSERVE                                                                 YES   NO
Backrest of chair is supporting person's back
Forearms are parallel to floor
Knees at angle of 90 - 110 degrees
Legs supported by chair seat
Person's eyes are 18 - 30 inches from screen
Top of monitor at or below eye level
Monitor viewed straight ahead with no neck turn
Wrists are in proper position (Wrist Support in use, or wrists held at level they
would be if there was a Wrist Support)

Typing stand (document holder) at same height and distance as screen [leave
blank if no stand in use]
No glare of outside light, overhead lights, or work lights on screen.

• Initial rate was 10% safe behaviors
• Plotted on Control Chart
• Pareto chart of individual items
established
• Safe behavior increased to 83% within 5
months, a significant increase
Barriers to Good Data Use
• Higher ups will use it as a ―hammer‖

• Subjected to quotas and targets imposed from above

• Fear (―accountability‖) used as a ―motivator‖

• Actions and Explanations as a result of random fluctuations

• Perceived loss of control over portrayal of performance

The Biggest Barrier = FEAR
Data Sources

• Worker and Customer Opinion
• Expert Review
• Process Measures
– Process Approach
Data Sources                                          THEORY

Dollars     Process
Hours                                                      OUTCOMES
Materials        Cycle Time                Product      Mission Progress
Data                                       Service      Commitments Met
In-Process                             Stewardship
Budget            Inventory                Schedule     Profit
Safety
Procedure                          Satisfaction
Compliance
Rework
Idle Time

Waste
More Fun with Numbers

•      Boston Herald (Dec 23, 1996):
"Murders Sink To 30-Year Low "
• "....major crime such as homicide is down.
Over the last two years there has been a
52% reduction in shootings..... that's a
dramatic difference....Police
Commissioner Paul Evans has succeeded
where others have fell short...."
More Fun with Numbers 2

Homicides per Year, Boston MA

Average = 88.5
160
(1976 - 1989)
140
Number of Homicides

c-chart UCL
120
100
80
60
c-chart LCL                        New Chief
40
in 1993
20
0
1964

1966

1968

1970

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996
Year
“Murder rate rises in Boston”

BOSTON (AP) — The number of murders in
Boston rose in 2000 after years of steady
decline, partially due to an alarmingly violent first
six months.
The final tally was 37 — up from 31 last year.
[Police Commissioner] Evans noted that there
were no homicides in October or November

For 2001, AP reported a 67% increase
Conclusion – Data Sanity

We can either react to numbers, with explanations
of every percent change, with the inherent
frustrations, fear, and failure
Or
We can understand our data, put it to good use,
and apply valid management principles

The choice is ours.

```
To top