The Effective Industrial Statistician: Necessary Knowledge

Report
The Effective Industrial Statistician:
Necessary Knowledge and Skills
William Q. Meeker
Department of Statistics
Center for Nondestructive Evaluation
Iowa State University
[email protected]
QPRC 2009
IBM, Yorktown Heights, NY
3 June 2009
1
Overview
•
•
•
•
•
•
•
•
Evolution of the Industrial Statistician
What Applications do Industrial Statisticians See?
What Tools Does an Industrial Statistician Need?
Statistics Graduate Program
Personality of a Statistician
Other Skills
Internships for Statistics Graduate Students
Concluding Remarks
2
Evolution of the Industrial Statistician
• Snapshot at 1975
• Snapshot today
Can we extrapolate into the future?
3
Typical Tasks
for an Industrial Statisticians in 1975
•
•
•
•
•
Design experiments
Modeling and analysis of data (including general number crunching)
Interpret results
Training
Conduct research for nonstandard problems
Many US statisticians worked in a statistics group within the company, e.g.:
Allied Chemical
Bell Labs
GE
IBM
Pratt and Whitney
RCA
Amoco
DuPont
GM
Kodak
Proctor and Gamble
Shell
How many remain?
4
The Industrial Statistician’s
Environment in 2007
• Modern statistical software can do an effective job of
modeling and analysis of data and designing simple
experiments, and readily accessible to all
• Statisticians tend to get involved in more complicated
interdisciplinary problems
• Training customers (perhaps increased due to six-sigma)
• Customers do not want pay for research (or even
technical reports)
• Fewer “Statistics Groups.” Most statisticians integrated
into product development or manufacturing groups.
• More need to be proactive, rather than reactive
5
What Applications do
Industrial Statisticians See?
• Product quality and manufacturing
–
–
–
–
•
•
•
•
Product design (including reliability)
Process design (including reliability)
Process monitoring
Warranty and other reliability field data
Marketing
Financial services
Environmental issues
Many other business processes
6
Some Statistical Tools Needed by
Industrial Statisticians
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
Bayesian Statistics
Categorical data methods
Censored data analysis
Design of experiments
Graphical methods
Image analysis
Multivariate analysis
Optimization
Regression analysis (linear and nonlinear)
Reliability theory
Response surface methods
Simulation
Spatial statistics
Statistical computing and programming
Survey sampling
Time series analysis
7
What Should Be in a Statistics
Graduate Program Core?
• At least two semesters of mathematical
statistics (probability and statistics, perhaps
stochastic processes).
• At least two semesters of statistical modeling
and methods with applications (linear and
nonlinear regression and maximum likelihood)
• SAS and R (or S-PLUS) use and programming,
plus exposure to Excel, JMP or MINITAB
• A creative project, thesis, and/or a course in
consulting, and corresponding internship
experience.
8
Which Statistical Electives?
•
•
•
•
•
Design of experiments
Statistical methods for reliability
Statistical methods for quality
Others according to interests
Important: While pursuing a graduate degree, you
cannot learn everything that you will need.
– The purpose of education is to learn how to learn.
– Statisticians should be prepared to learn (and in some cases
develop) new methods to meet the needs of the client (through
continuing education and self-study).
– In some cases statisticians may need to suggest hiring an
outside consultant for special problems
9
Personality of a Statistician
• The joke: A statistician is someone who loves to
work with numbers but who did not have the personality
to be an accountant.
Unknown
• The reality:
– Today’s Industrial Statistician works almost
exclusively in collaborations with scientists,
engineers, managers, and other non-statisticians.
– Interpersonal skills are extremely important
10
Other Skills of an
Effective Industrial Statistician
• Communications skills
–
–
–
–
Written
Listening
Presentation
Interpersonal
• Leadership skills (needed to be proactive)
• Knowledge of relevant subject matter areas, e.g.:
–
–
–
–
–
–
Biology
Business and Finance
Chemistry
Engineering
Genetics
Physics
• Flexibility and adaptability
11
Communications with Clients
• Statisticians should strive to learn some of
the scientific/engineering background in
the area of their client.
• It is imperative that the statistician learn
and use the language, notation, and
traditions of the client’s area.
12
Thanks to
•
•
•
•
•
•
Mentors at GE
Mentors at ISU
Colleagues and supervisors at Bell Labs
My students
My understanding family
Interesting/Helpful clients and access to
real problems
13
Internships for
Statistics Graduate Students
• Valuable experiences possible (not the same as
working in a university consulting lab)
• Projects may lead to professional society
presentations or publications
• Effectiveness is highly dependent on the kind of
project and attention of the mentor
• Exposure to the business environment will
provide perspective in subsequent years of
study and for the eventual job search
14
Concluding Remarks
• “Industrial Statistics” is nearly as broad as the
Statistics discipline itself.
• In spite of the new ability for others to do their
own data analysis, there will continue to be
healthy demand for statisticians in industry (but
in somewhat different roles).
• The truly effective industrial statistician will be
knowledgeable about the company’s business
and the science and engineering used there,
broad in perspective, and proactive in their work.
15

similar documents