Some postdoctoral fellowships work out well, and some do not. This talk reviews strategies for tailoring a successful post-doc experience that aligns with your career goals. The talk also gives some general career advice, and compares the pros and cons of different career paths.
2. 1. Plan
SAMSI is a social machine that was built for two purposes:
ā¢ Advance the interface of statistics and applied mathematics,
ā¢ Further the careers of its members.
I believe the ļ¬rst goal is being amply addressed, for lots of obvious
reasons. This talk will focus on the second purpose.
I intend to be speciļ¬c about how a SAMSI postdoc can be a force
multiplier for career advancement. And I shall connect that more
generally to strategic ways in which one can grow a career. The emphasis
is upon academic careers, but we shall talk about other paths too.
2
3. Chris Farley, as motivational speaker Matt Foley on Saturday Night Live
3
4. However, despite my complete lack of credentials in professional
development, there are a few reasons why I may be able to make this
worth your time:
ā¢ As a data scientist, I think I understand aspects of our profession
that aļ¬ect career growth.
ā¢ As a member of the ASA (and ENAR, IMS, ISBA, ISBIS, MAA,
and AAAS), I have a pretty good knowledge of the resources our
professional societies oļ¬er.
ā¢ As someone whoās worked for four universities and three federal
agencies, in managerial and submanagerial capacities, I have some
experience and a lot of war stories.
I have no guarantees. When thinking about successful careers, I am
reminded of G. K. Chestertonās explanation of why angels can ļ¬y.
āBecause they take themselves lightly.ā
4
5. 2. Strategy
Not everyone climbs high. Many who do should notāDeming told a
roomful of GM executives that they got there by doing the wrong thing.
There are many reasons why good people donāt rise. All careers have a
stochastic component. Nonetheless, long-term planning can help.
Like wavelets, multiresolution analysis is important. One needs a plan
for the week, a two-year plan, and a ļ¬ve-year plan. (These plans need
not be professionalāsometimes one must achieve personal goals ļ¬rst in
order to focus on a career.)
One needs to think two jobs ahead.
5
6. For almost everyone, advancement requires that you move. Often one
must be the bullet-proof best candidate before promotion happens.
In business or government, in order to be promoted internally, you ļ¬rst
need to train someone to replace you.
If you change jobs, try hard to leave only friends behind. Your greatest
asset as a data scientist is your reputation.
Through the ASA/IMS/MAA/AMS, there are many chances to make
friends at other universities, companies or agencies. Use these contacts
to learn of opportunities and to avoid war zones.
Think of your job as work, and your career as your hobby.
6
7. The kinds of strategies you need change with age. What makes you a
star as a young data scientist can doom you to mediocrity later on.
At ļ¬rst, one needs computational skills and good training. Later,
especially if one becomes a manager, the technical skills are less
important but organizational skills and oneās network are more critical.
Some general guidelines:
ā¢ The Law of Multiplication of Advantage (I. J. Good)
ā¢ The Peter Principle (Laurence Peter)
ā¢ Avoid Other Peopleās Nonsense (Lynne Hare)
ā¢ Donāt be Evil (Googleās SEC business statement)
ā¢ Honest Work is Never Wasted (Persi Diaconis)
ā¢ All That Matters is One Thing (Jack Palance, City Slickers)
7
8. āChance favors the prepared mind.ā
Eric Bogosian, playing the criminal genius Troy Dane in Under Siege 3.
(Also Louis Pasteur.)
8
9. 3. Tactics
āStrategy without tactics is the slowest route to victory.
Tactics without strategy is the noise before defeat.ā
Sun Tzu, The Art of War
9
10. 3.1 Postdoc Skills
Use your time as a SAMSI postdoc to:
ā¢ Learn how to do research (by yourself, collaboratively, and
cross-disciplinarilyāthese are all diļ¬erent).
ā¢ Learn how to teach: introductory classes, MS classes, and graduate
courses are all quite diļ¬erent too.
ā¢ Build out your computational skills. Nearly all young data scientists
will need to handle Big Data, and that takes āa very particular set of
skillsā (Liam Neeson, Taken).
ā¢ Give many professional talks, and ļ¬gure out how to improve each
time. Donāt be dull, and donāt lose your audience.
ā¢ Learn how to write a research paper. You want to leave SAMSI with
at least three publications.
10
11. ā¢ Practice consulting across hard and soft ļ¬elds. Data scientists need
this skill, and it is not as easy as some think. Communication is key.
ā¢ Write a proposal for research funding. It will need to be tailored to
whichever sponsor you targetāthey are not generic.
ā¢ Learn how to referee a paper. Ask to see the reports from other
referees who reviewed the same paper.
ā¢ Network professionally, both vertically and horizontally. Talk to the
graduate fellows, the emininent visitors, and to each other.
ā¢ Broaden within your ļ¬eld. Keep learning, and stretch yourself in
new directions as opportunities present.
ā¢ Learn about your ļ¬eld. Become familiar with the body of history,
wild character, in-jokes, and gossip that is our scientiļ¬c heritage.
11
12. 3.2 Use Your Societies
The ASA and the MAA provide:
ā¢ Salary surveys for statisticians and mathematicians, broken out by
covariatesāthis is key for negotiating job oļ¬ers.
ā¢ Professional job fairs, at every major meeting. Often the last day is
open to all, and it never hurts to look.
ā¢ Continuing education opportunities, as a student or a teacher.
ā¢ Sections, Chapters and committees that are ladders for leadership
within the societies, opportunities for networking with like-minded
people, and simple ways to build oneās resume.
12
13. ā¢ Practice in public speaking, and a 20-minute opportunity to advertise
yourself several times a year.
ā¢ Journals, a directory of members, on-line job listings, on-order
shortcourses, and so forth.
ā¢ Visiting lecturer programs, which provide good data science talks to
almost any organization.
ā¢ The opportunity to befriend and support worthy colleagues by
nominating them for committees or honors.
13
14. Professional societies are eager for you to deļ¬ne your own leadership
path:
ā¢ Jonathan Kurlander started the SIG on Statistical Volunteerism by
suggesting it in a letter to Fritz Scheuren, then the ASA president.
ā¢ Monica Johnston started a support program for M.S. level
statisticians by contacting the ASA Committee on Membership and
Recruitment.
ā¢ In 2009, Mingxiu Hu and Monica Johnston founded the ASA Section
for Statistical Programmers and Analysts, which ultimately led to
the creation of this Conference on Statistical Practice.
ā¢ John Bailer started Stats and Stories, a set of interviews with
statisticians about useful and oļ¬beat work they have done.
ā¢ And there are many other examples.
14
15. 3.3 Time Management
A sine qua non for success in any career path is good time management.
Some people are just not able to do thisāmostly teenagers, movie stars,
and pure mathematicians.
If you are vastly talented, then probably you can ļ¬nd someone to manage
your time for youāan executive assistant or an agent. Otherwise, you
must rely upon yourself (or your spouse). For career advancement, most
people need to manage their time themselves.
Jay Kadane once told me, when I was a junior faculty member at
Carnegie Mellon, that the key to success was to carry a pocket diary.
15
16. But there is more. Besides a pocket diary, it is good to have a make
a regular to-do list, and cross it oļ¬ as you proceed. Balance your
list to have a mix of easy and hard jobs, so you can feel a sense of
accomplishment even when tired or lazy.
Return phone calls and email promptly. A boss gets worried when an
employee goes darkāthat often means that a project is in trouble. Keep
the boss informed about delays in advance.
Everyone goofs oļ¬. Allow time for that, and donāt judge yourself harshly.
But sometimes one is tempted to goof oļ¬ too much. This can indicate
either:
ā¢ An immature personality, prone to computer games, or
ā¢ That your job is not suļ¬ciently challenging.
If the latter, replace the goof-oļ¬ time with systematic eļ¬orts to improve
your situation.
16
17. A main subcategory of time management is paperwork. Nobody likes it.
So, if you tackle it diligently and quickly, hitting the ball back over the
net every time it lands on your desk, you quickly get a reputation for
organization and responsibility.
No manager can conļ¬dently promote someone whose paperwork is slow
and problematic.
Work expands to ļ¬ll the available time. If you do a lot, you probably will
ļ¬nd yourself becoming faster, more decisive, and more focused. An A-
on four projects is often better than an A+ on one.
But good time management can make you start to feel like a circus
plate-spinner. Some people enjoy this, others are uncomfortable.
17
18. 3.4 Social Networking
It may seem ironical, but statisticians know all about this, at least from
a theoretical perspective.
Social network models in statistics began Holland and Leinhardt, but
were developed by Fienberg, Wasserman, Snijders, Handcock, Hoļ¬, and
many others. A simple version is:
logit pij = Āµ + Ī±T
i xi + Ī²T
j xj + Ī³T
ijxij + Ē«ij
where pij is the probability of a link from actor i to actor j, the Ī±i and
Ī²j are actor-speciļ¬c coeļ¬cient vectors for covariates xi and xj, Ī³ij is a
vector coeļ¬cient for dyadic covariates, and Ē«ij is random error.
18
19. What statisticians donāt know about social networks is how to use them.
David Krackhardt has done a study of Simmelian ties. These are triadic
links in social networks, as opposed to dyadic links.
Krackhardt found that in plays of Prisonerās Dilemma, people who had
only dyadic ties were about as likely to defect as strangers. But people
who have triadic ties are much more likely to cooperate.
Cliques provide a social context that deļ¬nes a team. So build a clique for
yourself. This involves:
ā¢ introducing people you know to others whom you know;
ā¢ organizing movie nights or dinners for gangs of friends;
ā¢ joining pre-existing teams and networking them together.
19
21. 3.5 Meetings
Many people look really stupid in meetings. My suggestions are:
ā¢ Stay focused; maintain meeting discipline.
ā¢ Be decisive, incisive, and concise.
ā¢ Make only a few main points.
ā¢ No meeting should last longer than an hour.
ā¢ Donāt think out loudāunless you are the smartest person in the
room you are guaranteed to look dumb to someone, and even if you
are the smartest you may still look dumb or annoy others.
ā¢ Watch the group dynamics and body language.
ā¢ Donāt pursue lost causes or raise dead issues.
ā¢ Try to act like the characters in West Wing.
21
22. 3.6 Sculpting Your Image
1. Sit in front, and ask one question.
2. If you go to a colloquium, look up the speaker beforehand and skim
the paper which is most pertinent to the talk. That way you will be
sure to have something smart to say, and you generate the illusion of
omniscience.
3. Read widely, especially in areas that are pertinent to your career path.
The history of math and statistics is good for all us. If you work for,
say, the FDA, you might also try Gina Kolataās Flu, and so forth.
4. Scan the newspaper (or a news website) every day. Find one article
that is a good topic for conversation with colleagues.
22
23. 5. Think of a new idea each week. Some of them will be good enough to
be the basis for a paper, or to pitch to the boss. The latter is a skill
worth practicing.
6. Learning to write well is a lifetime process. Cultivate a fussiness
about grammar, spelling, and condign expression.
7. Learning to teach is also a lifetime process. Practice it with an eye to
continual improvement. Perhaps you can volunteer to give a lecture
or a class on some aspect of quantitative science to people at work.
8. Seem happy and be active. Donāt complain (except rarely and
colorfully). Say nice things about people whenever possible. You
have to play politics, but respect the rules.
23
24. 9. Have some party tricks.
10. Reach out to a co-professional once a week. Call a friend from
graduate school, or a former colleague, or someone you met at JSM.
11. Regularly do something that invests in your career. Go to an ASA
chapter meeting, or learn a new software package.
12. Donāt use your hands when you speak. (I donāt know why, but a
management seminar told me it was badāmaybe it makes one come
across like Vincent DāOnofrio in Law & Order: Criminal Intent.)
A lot of a career is about acting. Put yourself in the place of your boss
or colleague, ļ¬gure out what they want, and then enact it.
24
25. āBe what you would seem.ā
Marcus Aurelius, The Meditations.
25
26. 3.7 Risk Assessment
You job is always in jeopardy. And any new job you take will have
unexpected pitfalls. So you should take some time to think about the
informal risk calculus that applies to your situation. You should always
have a sense of the probabilistic balance of costs and beneļ¬ts in terms of
your current job and possible alternatives.
As a general rule of thumb, your oļ¬ce is never more than two bad
managers (in time or hierarchy) away from a meltdown. And sane,
competent, honest managers are surprisingly rare.
If your oļ¬ce is in trouble, you can use that. The ļ¬rst people to leave
are the best (they tend to get good oļ¬ers). Join that wave. And when
the interviewer for your next position asks why you want to leave, the
answer will be obvious.
26
27. A key part of the cost/beneļ¬t tradeoļ¬ is getting accurate information. It
requires a combination of homework and scuttlebutt.
ā¢ Homework is researching alternatives, e.g., through the ASA or
SIAM jobsites. It also entails tracking management plans for your
group, trends in the ļ¬eld, etc.
ā¢ Gossip is where you learn which projects are hot, and with whom
it is good to work. It also includes salary. Americans tend not
to discuss salary. This code of silence advantages employers over
employees (e.g., gender discrimination in salaries, Lily Leadbetter,
oļ¬ce favoritism, Mrs. Parker and the Vicious Circle).
Factoid: Economists calculate that the average employee generates about
$66,000/year in proļ¬t for their company. Data Scientists probably
leverage more than the average.
27
28. ā¢ Avoid doomed projects, and ones that do not build new assets.
ā¢ Look for projects that cross departmental boundaries.
ā¢ Donāt be too risk averseāgamble when the payoļ¬ is right.(āThe
Lord hates a cowardā, Sean Connery playing Malone in The
Untouchables).
ā¢ Have multiple irons in the ļ¬re.
ā¢ Try to attend two annual meetings: your big professional society
meeting, for all the obvious reasons, and a smaller one in a speciļ¬c
domain, where you can grow inļ¬uential.
ā¢ Have an exit plan ready, and donāt wait when the weather changes.
At diļ¬erent stages of life, there are diļ¬erent constraints on the risk
analysis; e.g., children in high school, a spouse in a good job. If you
cannot relocate, then study local options and work to grow the career
potential of your current job.
28
29. Part of any risk assessment is a clear sense of oneās strengths and
weaknesses. Some categories in which you might rate yourself, say on a
Likert scale, are:
ā¢ public speaking
ā¢ computational skills (SUDAAN, R, SAS, etc.)
ā¢ analytic skills
ā¢ writing ability
ā¢ management capability
ā¢ leadership (this is diļ¬erent from management)
ā¢ cross-training breadth
ā¢ social skills
Sadly, some people tend to overrate themselves, especially in areas in
which they are most deļ¬cient.
29
30. āA manās gotta know his limits.ā
Clint Eastwood, playing Dirty Harry in Magnum Force
30
31. Steve Smaha, founder of Haystack Laboratories, said that people tend
to be pioneers, homesteaders, or farmers. (Of course, we understand
mixture distributions...) All three types can rise to the top.
His point is that people have diļ¬erent temperaments. One should
think about what kind of life one would enjoy, and make decisions
accordinglyāin other words, know your utility function.
Once a decision is made, donāt look back. āWhat is behind me, she
does not matter!ā (Raul Julia, playing the Italian racer in The Great
Gumball Rally).
To rise, one only needs to be the best in oneās group, not the best in the
world. From that standpoint, one strategy is to start in a large group,
establish credibility and contacts, then move to a small group where you
can be a star, and then perhaps move back to a large group with a higher
rank and accelerated experience.
31
32. 3.8 Publication
Even for non-academics, publication is helpful. The societies sponsor
many kinds of journals:
ā¢ Statistics in Biopharmaceutical Research
ā¢ Journal of Business and Economic Statistics
ā¢ Technometrics
ā¢ Journal of Agricultural, Biological and Environmental Statistics
ā¢ Journal of Quantitative Analysis in Sports
ā¢ Statistics Surveys
ā¢ Journal of Statistical Software
ā¢ Statistics and Public Policy
32
33. Henry Oldenberg was the ļ¬rst secretary of the Royal Society, and
invented the refereeing and publication processes. In his day, constraints
of time and paper and cost made that process necessary, but the Internet
has changed all that.
Henry Oldenberg. Heās caused a lot of trouble.
33
34. 4. Closing Soundbites
Your most important career asset is social capital. You build this by
helping others. Sadly, one can sometimes get more capital by sucking
up than by doing honest work. You have to consult your conscience on
where the right balance lies.
Try not to judge others harshly. Hermann Hesse wrote āWe most
despise those faults in others that we ourselves possessā
(Demian). We also despise faults that ļ¬atter our self-esteem.
Try not to hold grudges. A few years out, everything is less intense.
A Zen monk can be a CEO. Crafting sand mandalas and sweeping them
away is often good practice for corporate or government work.
34
35. Ethics matters, but good manners matter more.
Almost always say yes. Take risks, try new things, volunteer for more
work than you can reasonably handle. People have enormous capacity,
and the more you do the better you can get.
Donāt spam your colleagues with every minor accomplishment.
Be modest, or at least work hard to simulate it.
Donāt hang onāwhen the weather changes, leave.
Build skills (āYou know, like nunchuck skills, bowhunting skills,
computer hacking skills...ā, Napoleon Dynamite).
Use your SAMSI network!
35