“Learning more about specific lemur dietary needs will help us to focus conservation efforts on the best habitats that remain, to provide the best care for animals in captive breeding programs and to guide reforestation efforts," says Semel.
Customer Story
At Virginia Tech, collaborative research leads to a better understanding of species ecology
An interdisciplinary research effort sheds light on lemur behavior with the aim of improving conservation strategy and captive rearing outcomes.
Virginia Tech
| Challenge | Graduate students in statistics need experience applying their knowledge to practical situations. Graduate students in other disciplines need help making their research more statistically robust. |
|---|---|
| Solution | Virginia Tech’s Statistical Applications and Innovation Group (SAIG) pairs graduate-level statisticians with scientists in other fields to address complex research questions collaboratively. SAIG consultants use JMP® to make statistical analysis more approachable and interactive. |
| Results | Two Virginia Tech doctoral students, a statistician and an ecologist, explored out-of-lab nutritional data in JMP to identify biologically meaningful variables that help explain why wild lemurs consume soil. Their results aim to inform conservation management strategies affecting this critically endangered species. |
Scientists + statisticians = discovery. It’s simple math. So says Jiangeng Huang, PhD candidate in statistics at Virginia Tech.
And Huang would know; over the past three years, he has collaborated with researcher Brandon Semel, a PhD candidate in fish and wildlife conservation also at Virginia Tech, to explain the ecological function of soil consumption in wild lemur populations. Their combined efforts have revealed a number of biologically important variables representing an expansion of scientists’ understanding of lemur behavior.
From fieldwork to robust statistical modeling
Primates have long been known to consume soil – a behavior known as geophagy – says Semel of his research; however, scientists have yet to explain why this behavior evolved and what purpose it now serves. Existing hypotheses take into account a combination of nutritional, geographic and demographic factors and range from parasite mitigation to mineral supplementation to flushing toxins. Cracking this question could be the key to understanding how conservationists can better direct their management decisions to increase the likelihood that the species will survive.
“Forests in Madagascar are rapidly disappearing, threatening many of the over 100 lemur species that depend on them for food and shelter with extinction,” Semel says. “Learning more about specific lemur dietary needs will help us to focus conservation efforts on the best habitats that remain, to provide the best care for animals in captive breeding programs and to guide reforestation efforts.”
Lemurs, including diademed sifakas, the species that Semel studies, are endemic to Madagascar. Each year, Semel makes the trek to the nation’s eastern montane forests to observe lemur behavior and record data such as bite rates, food quantity, food type and the time spent consuming different foods and soil. This data then gets converted to nutritional information that can be more easily studied in the lab: minerals, fat, calories, fiber, protein.
Returning to Blacksburg from one such trip three years ago, Semel found himself facing several data challenges for which he was ill-equipped; the data set was large, the variables multicollinear and the missing values many. In order to augment the quality and precision of his findings, Semel sought a partnership through the Statistical Application and Innovation Group (SAIG), a Virginia Tech Department of Statistics initiative.
PHOTOS BY BRANDON SEMEL
A university initiative provides applied experience, better science
Offering statistical collaboration, consulting and support for research scientists in other disciplines across the university, SAIG is the embodiment of the scientists + statisticians = discovery formula. The group’s objective is to assist researchers in designing more robust experiments, honing their data modeling and analysis skills, and teaching them to use the software they need to carry any newly acquired statistical skills with them into future research. In return, graduate students in statistics gain an opportunity to apply their expertise to real-world scenarios – valuable experience for any new graduate looking to pursue an industry career.
Since beginning his studies at Virginia Tech – and joining SAIG – Jiangeng Huang has worked on more than 60 projects with researchers from fields as disparate as forestry and civil engineering. Some were quick-turnaround projects; others, such as his work with Semel, much more extensive. “We listen to and learn about our colleagues’ research context and then figure out the most appropriate algorithms to use to solve their problem,” Huang says. “We just teach them how to fish, and they can then fish by themselves.”
When Semel first turned up at SAIG, he knew a bit about fishing; that is, he knew about linear models but not much about how to apply them. Huang, on the other hand, knew nothing about lemurs. That soon changed and the two have now become trusted colleagues and good friends. Collaboratively, they make a formidable duo.
“I was a little worried that the people at SAIG and I would just talk right past one another from our areas of expertise… and that it would be a challenge coming to a mutual understanding,” Semel recalls. “I had looked over a lot of statistical approaches to the problems that I was having, but felt I needed help applying them to my particular set of problems.”
JMP® simplifies the stats
After sitting down with Semel to better understand the goals and parameters of his research, Huang began exploring the data. Together, they hashed out an approach which eventually led them to build a series of regularized linear models that allowed them to, in one step, select important features and estimate their effects. By combining lasso and ridge regressions with the elastic net, they were able to address the multicollinearity issue. And to build these models, they used JMP.
“I like to use JMP because it's so interactive,” Huang attests. “It's very user-friendly, and there are so many tools available.” Also, he says, because users don’t need to learn to write code, JMP makes statistical methods more approachable for scientists like Semel who need to free up time to focus on fieldwork. “Researchers from other disciplines all have different levels of programming skill and statistical knowledge. Many ongoing research projects are interdisciplinary in nature and require effective collaboration from both domain expertise and statistical computing,” Huang says. And JMP makes this process easier. “Especially at the early exploratory stage, we want to try many different methods to figure out which one works best for the problem at hand. I can't spend 10-plus hours coding things from scratch for every potential method we want to try, when I am also working on multiple research and consulting projects at once.”
That’s why the SAIG initiative aims to not only equip researchers with a knowledge of the basic principles, but to teach them how to use JMP. With the wide-ranging prepackaged options available in the software, researchers don’t need any other tool, Huang says. “JMP makes it so easy for researchers to analyze their own data. I think that's the really cool thing about JMP.”
-
-
Each year, Semel makes the trek to the nation’s eastern montane forests to observe lemur behavior and record data such as bite rates, food quantity, food type and the time spent consuming different foods and soil.
PHOTOS BY BRANDON SEMEL
A generalized linear model resolves the challenge of zero inflation
Semel has an especially complex data set. Because zero values are recorded in any instance in which lemurs are observed eating food but not soil – and not all lemurs eat soil – this means that as much as half of any given data set can be zeros. “This data set has a special structure in its response variable,” Huang says. “We call it zero-inflated. If you look at the response variable, there are counting data with many zeros. It's a non-normal data set, and we handled that in JMP by extending to a generalized linear model using our link function. We also need to be creative with the link function and in this case, we found a zero-inflated Poisson model works well.”
Dealing with “non-normal” data sets can be a challenge even for a trained statistician. But knowing that this scenario would certainly not be the last in Semel’s career in the field, Huang worked with him to develop a standard method that would be easily replicable and adaptable in JMP. “You can do this yourself,” Huang recalls telling him. "You can do this in JMP.”
Experience that prepares for the world ahead
“I believe this collaboration trend is a very strong part of our program,” Huang says of SAIG. “Statistics is very interdisciplinary and SAIG allows us (statisticians) to see how these statistical methods are applied to real problems. How do we actually use these algorithms? I might be familiar with them now, but if I don't apply them… [then my skills could get rusty].”
It all comes back to the collaboration paradigm, he says. “You advance research, but you're also learning in this process; you're learning skills that you can apply when you move on to do something that's not pure statistics. And the research itself is interesting. Working together, you come up with a solution.”
While Semel continues to wrap up his research on lemur geophagy, he is now looking at how lemur populations will be affected by climate change. “We know that some forest types can support more lemurs than others,” he says. “We also know that climate change is already causing shifts in forest types around the world. By studying lemur diets and populations across a range of forest types, I am hoping to predict future lemur abundance based on current protected area coverage and various climate change scenarios. That modeling will require many more trips to SAIG!”
Adds Huang: “We all have different talents. That's why I like to collaborate. I chose statistics because I can play in everyone's backyard.”
The results illustrated in this article are specific to the particular situations, business models, data input and computing environments described herein. Each SAS customer’s experience is unique, based on business and technical variables, and all statements must be considered nontypical. Actual savings, results and performance characteristics will vary depending on individual customer configurations and conditions. SAS does not guarantee or represent that every customer will achieve similar results. The only warranties for SAS products and services are those that are set forth in the express warranty statements in the written agreement for such products and services. Nothing herein should be construed as constituting an additional warranty. Customers have shared their successes with SAS as part of an agreed-upon contractual exchange or project success summarization following a successful implementation of SAS software.