An Interview With Bob Derr, Senior Research Statistician Developer
By Maura Stokes
Bob Derr is a Senior Research Developer with the Linear Models Department at SAS. He began working
here in 1994 while he was a student in the Statistics Department at the University of North Carolina
at Chapel Hill, and he accepted a full-time position in 1998. Before his move to North Carolina, Bob
had received a master's degree in mathematics with a concentration in statistics at Indiana University.
He then worked as a statistician at the Bureau of Labor Statistics in Washington DC for three years.
Bob supports the CATMOD, LOGISTIC and MULTTEST procedures, in addition to providing leadership in the ODS Graphics
development project. He implemented newer techniques, such as exact logistic regression, in PROC LOGISTIC, and he
also programmed the adaptive p-value adjustments in PROC MULTTEST.
What got you interested in pursuing a PhD in statistics?
I decided that I didn't know enough statistics when I was working at BLS. I was interested in the work,
but I wanted a stronger background. I'd taken a year of statistics at Indiana, and I was taking additional
graduate courses at George Washington University. But it wasn't enough for me.
Why did you pick UNC-Chapel Hill?
I chose UNC because it had such a theoretical department - I was serious about learning
a lot more about statistics - and I was aware that the Research Triangle area had a lot going
on statistically, with the three universities, pharmaceutical companies and government agencies.
What was the value of working at SAS as a student?
It gave me great exposure to the software development environment. I tested an early
version of SAS/INSIGHTŪ software, and I worked on SAS/STATŪ documentation. By the time that
I was finishing my PhD program, I had had experience in academia, government and private industry.
When I received the SAS job offer, I was very sure that I wanted to pursue that route.
What's a typical day in development for you?
First I check my e-mail to make sure I didn't destroy the previous night's
builds. Usually all is well, so I can start working on a project. Usually,
I'll have thought about a problem from the previous day (for example, when I'm driving
or taking a shower), and I often have a new idea about how to attack it by the time I get
back to work. So, I'll give that a go for a couple hours and see how much progress I make.
At some point, I'll need to take a break, and I'll often move to another coding project.
Generally, I'll have one major project and several minor projects going on at
the same time. And if I need to take a break from coding, I'll pick up a journal article
and turn my attention to that.
On a perfect day, I'll also have lunch with my youngest daughter, who is at SAS day care.
Your day can't be that much under control!
Well, I do have a lot of days that are a lot like that. But of course,
there are those days when I get e-mail about problems from my testers or a question from
Tech Support that needs a prompt reply, and then my day can take a very different direction.
What's your reaction when a tester finds a problem? I do hear screams occasionally from my office.
That's not me! I really like to hear about the problems. It means more work for me, of course,
but it also means that my software is getting a good workout before it goes to customers. I
get really nervous when new software testing isn't generating a number of problems, because I
worry that it's not getting the right attention.
What personal strength helps you most as a developer?
Persistence. I've always stuck with something until I get it right, and I find that especially helpful
in my programming. I'm often implementing new algorithms, and my first pass often raises issues
I hadn't anticipated. It really takes a lot of persistence to stay with it, work through the challenges,
and make the code succeed. Also, my background in theoretical statistics means that I can handle any statistical
paper that comes my way (I hope).
I know you played a major role in ODS Statistical Graphics Development. What did that entail?
As the ODS graphical work came together (this was work done by people in the graphics, Base,
and statistics area), I was one of the first developers asked to implement it. We often do that
with new techniques - have a few developers act as pioneers and then serve as advisors to the
rest of the development staff. I became a member of what we called the ODS Graphics Junta, which
reviewed every single statistical graph that was added to SAS. We wanted the graphs to represent
our best practices and to be internally consistent. Largely, I think we succeeded.
You mentioned reading journal articles. Is that how you keep up with methodology in your areas?
That's one way. The statistical developers are responsible for keeping up with the areas they support.
I do keep up with various journals, and every so often I'll do a term search in CIS (Cumulative Index
of Statistics) and see what I've missed. I'll attend courses on new methodology at conferences like
JSM, and I also attend sessions in my areas. And sometimes it's not the talk that's valuable to me.
It might be an off-hand remark made by a discussant, or a comment made by someone sitting in the audience,
that gives me a clue that I need to follow up on a new idea or technique. I keep in regular contact with
various experts in the field.
And of course, there are a number of excellent statisticians here, and we trade ideas and
connections all of the time.
Do you ever get good suggestions from users?
Sure. A number of suggestions come in from Tech Support. Sometimes a user asks for a
technique that we don't offer, and sometimes a user has a good idea for a new option or
feature to add to an existing procedure. Everything gets passed on to the development staff.
So users should feel free to communicate with Tech Support about new features?
Right. Even a quick e-mail gets recorded. It also helps if users can provide a
citation for the method. Sometimes it might seem obvious that a new methodology would be on
our radar, but having specific user requests is very valuable. It helps us set priorities.
Bob, you also mentioned that you worked on documentation as a student. Did that give you an
advantage when you started to work as a developer?
Absolutely. All SAS/STAT developers write their own documentation, with a homegrown system based on LaTeX,
which I had already used for several years as a student. Also, I learned how long it takes to write good
documentation and maintain consistency across an entire book. So, as a developer, I write documentation
as I write new code. When it's time to finalize the documentation, I revise my new sections many times before
I even send them out for review and editing.
We've put you out on the road recently to present the logistic regression tutorial. How do you like that?
I really enjoy it. I have to say that I get nervous the half hour before the course begins.
But once I get started, I'm fine! I find it really challenging to get statistical concepts across
to a variety of SAS users, and it's a lot of fun to explain how to use the software I developed.
I understand that you actually gave two seminars at this year's SAS Global Forum!
Well, I gave the tutorial, which goes for two hours, and then I moved into the hall to follow up on some questions
from the attendees. They had several interesting problems in logistic regression, so we found an empty meeting room
and had quite a long discussion. By the time the last person left, two more hours had passed!
I know your family takes top priority these days, but weren't you a member of the infamous Elvis on First softball
team that won the SAS intramural championship years ago?
Oh gosh, yes. And so were you. We won?
I can't remember what position I played.
I was outfield and easy out. I do remember thinking that so many statisticians on one team should be illegal.
We were particularly helpful to the umpire in tracking balls and strikes.
I'm not sure they saw it that way.
So, Bob, almost fifteen years at SAS. What's the best thing about working here?
Definitely the people. It's great to be surrounded by so many excellent statisticians. It's an
atmosphere that really promotes learning the new statistical methods and computational techniques
that I need to do my job. I never run out of challenges.
