University Times

Researchers are in the business of asking questions in order to further knowledge and increase understanding in their field.

However, to produce meaningful results, experiments must be designed properly and the results analyzed correctly.

That’s where statisticians come in.

“There’s tremendous demand for statisticians,” said statistics professor Allan Sampson, who noted there also is an increasing emphasis on the need for statistics education in a variety of disciplines. Most arts and sciences programs at Pitt now either require or strongly recommend that students have at least one statistics course, he said.

“A well-designed experiment saves money over a failed experiment. It also costs less money if you’ve designed it well,” he added. “You can save big money with a clever design” that may enable a researcher to decrease the number of subjects needed for a particular study.

“For most statisticians who are in a collaborative setting, their most valuable contribution is helping to design studies so that the person who’s conducting them has the possibility of getting the results they’d like to see,” Sampson said. “Nobody wants to spend a few hundred thousand dollars for a study and discover ‘I didn’t run the right study, I can’t answer the question.’”

Members of the Pitt community in need of assistance have a free resource in the Statistical Consulting Service (SCS). Staffed by graduate students from the Department of Statistics under Sampson’s direction, SCS offers expertise in designing experiments and interpreting results. While it proves to be an invaluable resource for faculty and students University-wide, SCS’s main goal is to serve as a training center for the department’s graduate students, Sampson said.

Of course, not all students are cut out to be consultants, but many will end up in collaborative roles once they finish their studies. They may work in university settings or in corporate environments ranging from finance to pharmaceuticals — situations in which they will be called upon to work well in partnership with others.

Pitt’s applied statistics master’s program and statistics PhD students work for a term as SCS consultants as part of a required course. “The primary purpose is to make sure graduates of our program have experience. In one semester they truly understand what consulting collaborative work is for a statistician,” Sampson said.

“What I really want to do is give them, as much as I can, the closest experience to working in a non-university setting inside of graduate school. It is a challenge for them because I push pretty hard in that regard to make it different from a class.”

The course typically has four-six students each term, which means they can take on 20-30 assignments. Time constraints preclude students from consulting on extremely complicated projects, but the range of experience they gain is huge, he said.

SCS clients must have some connection to Pitt, Sampson said, adding that they come from all areas, ranging from undergraduates doing honors theses to graduate students writing dissertations to researchers who lack funding for or access to a statistician’s help.

“There is more demand for services than there are people to provide that applied help,” he said. Recent projects have included determining what factors most influence response rates in email marketing efforts, describing variability in forest plant life and analyzing data that may result in new and better ways to identify lymphoma cells.

“The variety is enormous,” Sampson said.

Everett Ramer, a developer/demonstrator in the Department of Physics and Astronomy, recently called upon SCS to help him approach the problem of measuring the lifespan of a subatomic particle that decays rapidly — in a matter of microseconds.

“You need to fit the data to an exponential curve, as opposed to fitting it to a straight line,” he said.

Ramer met with consultant Sung Lee to discuss the problem and present the data. They met again later to go over ideas for contemplating the puzzle and, at that final meeting, Lee presented a walkthrough of how to model relationships between the data Ramer provided. Although Lee wasn’t an expert in particle physics, Ramer said he was well equipped to present some ideas for dealing with the data. “He came up with a very good suggestion,” said Ramer, who was pleased with the consultant’s guidance.

Will Brogan, a fourth-year PhD student in biological sciences, also praised the consulting service for the one-on-one assistance he received to aid his research into factors that may decrease the effects of contamination on aquatic systems.

SCS had been recommended by others in his lab and department, said Brogan, who admitted that although he has studied statistical methods and has a basic understanding of how to test different types of questions, he is not an expert.

He said his work is at the edge of what has been done previously, so no templates for the experimental design existed.

He said he met six or seven times with SCS consultant Carolyn Popplewell. “She was great,” he said, commending her ability to understand the questions being addressed although she is not a biologist.

He said she took detailed notes on the design and quickly picked up on issues surrounding the interactions among many different organisms.

Brogan said his goal in seeking the consultation was to verify that what he intended to do with the data was correct and to determine whether there were better ways he could deal with the information.

“It felt good to get the confirmation,” when she agreed that the test he was considering was correct.

“Having someone sit with me through six or seven sessions was really awesome,” he said, adding that although he’d also discussed the issue with his advisers, he would have felt he was imposing on them had he wanted to meet that many times about the experiment design.

“She helped me all semester” with two projects in meetings conducted at the Center for Statistics office in the Cathedral of Learning. “It’s a nice resource when they’re intimately knowledgeable about your projects,” he said.

Students don’t face clients unprepared. The course begins with instruction on ethics — both the ethics of science in general and ethics as it pertains to a statistician’s work — and discussion about the skills involved in consulting. Sampson said initial class meetings get students thinking about how to present themselves in a professional manner and cover interpersonal skills such as active listening and reading a client’s body language.

Sampson pre-screens clients and pairs them with students who, depending on the complexity of the projects, may take on four-six clients over the course of the term.

After being paired with their clients, students set up an initial meeting to learn about the client’s project and needs.

They then gather their ideas and make a presentation to their classmates on how they intend to deal with the statistical issues involved.

“They present the case, the science involved, what the client wants to accomplish, and their beginning ideas on how to deal statistically with the project,” Sampson said. Classmates then are free to ask questions and offer suggestions during class sessions that meet three-six hours per week.

Sampson also meets privately with students and exercises quality control over their work. Although students have between a year and a half and three years of graduate work under their belts by the time they take the course, they typically need encouragement until they gain confidence in their consulting skills.

Sampson said he aims to be tough yet supportive to help students make their transition into the working world less of a shock.

“My goal is to make it tough. In business there’s no excuse, ‘My dog ate my homework,’” Sampson said. “If you make a commitment to the client, you meet the commitment. In a job, you’ve got to meet deadlines.”

Sampson takes pride in giving students a bit of experience in a sheltered environment before they step out into the real world. They frequently have the opportunity to discuss details of their SCS experience in job interviews, where employers are eager to learn how well they work in a collaborative environment.

“Every student includes this on their resume as practical experience,” Sampson said.

And he loves to watch them transform into capable collaborators as the term progresses. “They come in and they’re nervous — they can’t organize, can’t present. By the end they can go up to the whiteboard, make confident presentations and understand how to handle themselves in this regard.”

Sampson said he makes a point of having students focus on their feelings at the beginning of the semester and compare how they feel at the end so they recognize all they’ve gained. “They have amazing growth as an individual and aren’t so timid,” he said.

The consulting experience also teaches them how to teach themselves when a client’s project goes beyond techniques that were taught explicitly in class.

“Consulting is like taking a core sample — they don’t know where they’re going to be,” Sampson said. “They have to combine everything they’ve learned — little bits and pieces — in a new way to handle a problem.”

In their coursework they’ve learned various methods, but in consulting they must decide on the appropriate ones. “They’ve got to start thinking vertically rather than horizontally,” Sampson said. “I think that is the biggest challenge for them.”

Sometimes the work involves analyzing data; other projects are related to designing a project or honing a research proposal.

Often, the student consultants won’t see the end result of their efforts. But sometimes they are named as co-authors on published research, or are asked to attend PhD defenses for candidates they assisted.

“It’s rewarding for the consultant to see what they’ve done here play out and really have value to it,” Sampson said. “We do this and people actually benefit from it.”

Sometimes the experience leads to an entirely new career plan for the consultants. Kaleab Abebe, who started in the statistics department’s master’s program in 2004, intended to complete his degree then work in industry. His experience in the consulting course, in conjunction with Sampson’s influence, prompted him to go on to complete his PhD — a path he once equated only with teaching and theory-driven academic publishing. He now works in the School of Medicine.

“My mind was closed before taking the course and talking with Dr. Sampson,” said Abebe, who since 2009 has been an assistant professor of medicine, biostatistics and clinical and translational science, and a statistician in the Center for Research on Health Care Data Center.

Most of his work involves collaborating with University investigators on projects and proposals, but he also teaches and does some research of his own. “The PhD was a good route to take,” he said.

His consulting experience in grad school helped prepare him for the variety he faces in his work today. The data center’s consultants are paired with investigators based on their interests — he favors trials and longitudinal data analysis — and their workload.

“It was a good way to get exposed to a whole lot of different projects,” he said.

Abebe has seen his grad school experience come full circle in his work at the data center. His first project as a student consultant was with psychiatry faculty member Rebecca Thurston, who was doing research on hot flashes. “I was extremely nervous working with a full-fledged investigator when I was only in grad school,” he recalled. He recently had the opportunity to work with her again, this time as a fellow faculty member collaborating on a grant proposal.

Abebe recalled Sampson’s emphasis in class on the importance of interpersonal skills  — lessons that have paid off in his professional life. He said his consulting experience taught him how to discern when a relationship with a client is not going to work out and how to communicate with investigators in a way that tactfully gets his point across. “Half the time it’s not what you know statistically, it’s more about working with people,” he said, noting the importance of being able to communicate statistics-speak in terms non-statisticians can understand.

He said Sampson’s emphasis on collaboration helps statisticians to be viewed not merely as a utility — someone who runs the numbers and produces the investigator’s desired results — but as a valuable collaborator.

“The better we work together from the beginning, the better it’s going to be,” Abebe said.

Information on the consulting service is available at www.stat.pitt.edu by clicking on “Statistical Consulting Center.”

—Kimberly K. Barlow