Leading a SAMSI Working Group: One Size Does Not Fit All

The following was written by Paul Brooks, Associate Professor, Department of Statistical Sciences and Operations Research, at Virginia Commonwealth University.

I came to SAMSI to get “real” statisticians to look at data that my microbiologist/clinical colleagues were generating to discover new patterns related to human health. Whenever I work with colleagues in the biomedical arena and they learn that I am a quantitative person, they automatically assume that I am a statistician and can help them with all of their statistical needs. They lump all of us applied math people together (I was happy to return the favor in my previous sentence). The experience depicted in this video rings all too true:

I watched this with some of my colleagues and we laughed until we cried because we have had nearly the same conversation several times over the years.

Last year, we were working on a grant proposal that asked for outreach to the community to make data available to analysts and train people on best practices. I had heard of SAMSI from a statistician in my department who had been urging us all to get involved. Then I discovered the \research program” concept that SAMSI has and further, that a theme for this year was Beyond Bioinformatics. The stars aligned. After speaking with Snehalata Huzurbazar, Deputy Director of SAMSI at the time, we agreed that I could lead a working group during the Beyond Bioinformatics program. Several other things fell into place, and I am now a visiting research fellow at SAMSI and leading two working groups.

Coming into the year, I was excited about the concept of a working group adding all kinds of new expertise to our ongoing projects. But I am a bit of an anxious person. Perhaps a nicer way to put it is, I am a planner. I like to have my stuff together well ahead of time. I met some SAMSI postdocs over the summer at a conference and asked them for advice on a successful working group. They said it takes a leader to assign tasks and hold people accountable for getting things done. Okay, I would need to learn how to do that.

When I arrived at SAMSI, I attended the opening workshop on ecological modeling and spied on them and how their working groups formed. Some of the working groups that formed there were huge! There were 35+ people in most groups, and many people were in multiple groups. How was I going to find enough ideas for that many people to work on? One group seemed to have a pretty good approach: they first brainstormed general topics of interest, then they collected whose to form article topics, then they began to work on what their paper titles would be, and even discussed target journals. That seemed pretty good for the first two days. I would try that. But of course it didn’t go quite as planned. Around 200 people attended the Beyond Bioinformatics opening workshop, and there were about 10 working groups proposed. The working groups formed after lunch on Thursday, but many of the people left right after lunch. The initial working group meetings were not as big as anticipated, and many of the working groups met in the same room.

working group meeting

MCDC group meeting during the Beyond Bioinformatics workshop.

My proposed working group, MCDC (Microbiome Community Dynamics and Complexity), was sort of split into two and one was merged with another working group and would be complementary to one or two others. Plans went out the window. One of the working groups started with about 5-6 people on site and about 25 people attending online via WebEx. For the first few meetings, we agreed on papers to read, and someone would lead a discussion of the analysis methods used. For a while, it seemed like only 2-3 of the 20-30 in attendance participated in the discussions. Some of these discussions were somewhat heated because of disagreements about appropriate modeling strategies. Perhaps people new to the held and graduate students/postdocs did not feel that they could weigh in on the issues yet. But they kept coming back.

The meetings reminded me of my early experiences with microbiome conferences. When I first started attending microbiome conferences, there was a group of postdocs who would microblog every slide from every presentation. They would describe what the speaker was presenting, then offer their opinion. They often disagreed with the presenter and with each other. Observing these discussions was incredibly helpful to me to understand a new field and to understand what the big questions were. Perhaps that’s the kind of service/entertainment we were providing with our working group meetings.

Fast forward to today. We still have 15-25 people at each meeting. Many more people are actively participating in the discussions than when we started. We have some subgroups who are working on different parts of a paper that we hope to write together. And we are laying the foundation for two additional papers to write together.

Each working group is a unique experience. Be flexible and know that different approaches to leading a working group can lead to fruitful collaboration.


SAMSI Paleo Family’s Secret to Staying Together for Many Years

The Paleo Family

The Paleo Family (minus one) (Left – Right): Martin Tingley, Liz Mannshardt, Bo Li, Peter Craigmile and Murali Haran.

Sometimes groups are meant to be. Take a working group that focuses on Paleoclimate data, who now affectionately calls itself the “Paleo Family.” This group started over five years ago at a program on Space-time Analysis for Environmental Mapping, Epidemiology and Climate Change at SAMSI’s opening workshop. Usually most groups dissipate after the research program is formally over, but some are still very active even several years later.

The group had started out as a larger group, but has whittled down to six people including former SAMSI postdocs Bala Rajaratnam, (now an Assistant Professor at Stanford University), Martin Tingley (now Assistant Professor at Penn State University), Elizabeth Mannshardt (now a Postdoc at North Carolina State University) and Murali Haran (Associate Professor at Penn State University) along with Bo Li (Associate Professor at University of Illinois at Urbana-Champaign) and Peter Craigmile (Associate Professor at The Ohio State University).

The group focused on the science of reconstructing past temperatures based on proxy records such as tree rings, lake sediments and ice cores. Modern instrumental records of weather only date back to about 1850, so proxies can help extend records to at least another 1,000 years.

The group has published “Piecing together the past statistical insights into paleoclimatic reconstructions,” in the 2013 Quatenary Science Reviews 35, 1-22; which is also referred to as the “monster paper,” by the group. This paper has been referenced in the Intergovernmental Panel on Climate Change (IPCC) report. This paper outlines the statistical challenges of using instruments and proxies to reconstruct past climate, in space and time, and lays out a flexible, Hierarchical framework for the reconstruction problem. Martin says, “Our goal was to demonstrate that more sophisticated statistical techniques allow the data to be used to address more scientifically interesting questions.”

Subsets of the group have published a number of relevant papers including, “Statistical modeling of extreme value behavior in North American tree-ring density series.” in the 2013 Climatic Change, 117, 843-858 and Tingley, Martin P. and Bo Li, comments on “Reconstructing the NH mean temperature: Can underestimation of trends and variability be avoided?” by Bo Christiansen. Journal of Climate 25: 3441-3446, 2012, and they have contributed to discussions of other paper. The group is in the middle of preparing another peer-reviewed paper. Martin says, “An important goal of paleoclimate science is to use the longer time frame afforded by the proxies to provide out-of-sample assessments of the climate models that are used to project future climate. We are building hierarchical statistical models to rank simulations of past climate based on their agreement with proxy-derived reconstructions.”

The “monster paper” was also used as the basis for a paleo climate seminar course at The Ohio State University (OSU) in 2012, “Statistical Methods: Paleoclimate Data and Models”. Martin and Peter gave lectures there, as well as Mark Berliner, professor of statistics at OSU and Jason Box, visiting scholar at the Byrd Polar Research Center and Professor at the Geologic Survey of Denmark and Greenland.

Bala Rajaratnam (L) and Elizabeth Mannshardt (R)

Bala Rajaratnam (L) and Elizabeth Mannshardt (R) recently visited SAMSI.

“I don’t think there had been something previously that laid out statistical considerations or methods when you are working with paleo data,” remarked Elizabeth.

Each person in the group comes from different research backgrounds so each person brings some unique skill to the group. Also, diversity is another thing that we are proud of the group. We come from six different countries across four different continents, so our discussions and research are the integration of strength from different cultures.

Martin helped set the agenda early on because he had the background working with paleoclimate data. Martin said, “This group has been an incredible opportunity and resource for me — five statisticians who are willing to learn enough science and teach me enough stats that together we make progress. I view my main responsibility as keeping the group engaged by hunting down interesting problems.”   The statisticians work in a range of fields from applied to methodology, and including theoretical statistics research.   Each brings different strengths and insights to the group. “Peter is the king of code. He programs faster than I can type!” said Elizabeth.

“The group has been very supportive to each other, even on other professional items such as helping each other out when someone is job hunting,” noted Bala. “It is also entertaining to see how long it takes six Ph.D.s to all get on Skype at the same time too,” quipped Elizabeth.

Our group formed a special session on paleoclimate at the Purdue Symposium in 2012. “I was thrilled to have my friends gathered in West Lafayette, and had discussions in person,” said Bo.

And what, exactly, is the secret to their success? “We picked a topic that was very interesting. We are all open-minded. We like to tease each other and everyone is fair game, and we even poke fun at ourselves. I think we really genuinely like each other,” said Bala.  “At the beginning of each semester, we all get together to figure out when we can carve out the one hour a week to meet with each other. We all still consider this very important”, said Elizabeth.

Thanks to a comment on a climate debate blog, the group has their rock band name: “Tingley and the Flamboyant (Professional) Statisticians!”