Chicken-or-egg: Which came first, the question or the model?

The questions we ask

The quest for generality is at the core of ecology. Once we observe some relationship between X and Y, we usually want to know how general that relationship is, and how much the relationship can vary in different conditions – whether it’s ecosystems, species, individuals, sites, or other groupings. In other words, we are interested in separating the variation we observe into (1) a general relationship between X and Y that explains what we see most of the time, and (2) conditions under which this general relationship may differ. These conditions can be ecological (i.e. maybe species differ in the way they respond to X), but they can also be due to the process of observing the relationships (i.e. maybe X responds to Y in site A as expected and B but not C and D, for reasons we haven’t measured).

As an example, we know that plants generally grow faster when they are exposed to more light. If we go out and measure plant growth rates in sites A, B, C, and D, we may see that this general relationship holds. But, some species may actually grow better in low light, and their relationship with light will differ. 

When we monitor biodiversity change, we are often asking a similar question. How are species’ abundances generally responding to global change: are they declining? stable? Growing? Once we know this general response, we also need to know how individual species are doing: which species are declining the fastest, and need immediate conservation interventions? Which ones are stable but not growing, but should continue being monitored in case this changes? Which ones are growing, and may be doing better after conservation or may not need any conservation action for the time being?

So, our question is generally framed in this way: what is the general relationship, and when and how much can this relationship differ between species?

How we answer these questions

To answer these questions, we build models, collect data, and evaluate how well our model explains the data. For simplicity, let’s explore an example where we want to understand Y ~ X for many species. One approach is to model Y ~ X for each species and then try to put all these models together to get some kind of average or agreement between them. Maybe we could look at how the slope of our models vary across these species, to make inferences about how the strength of the relationship between X and Y differs between species.  

Another approach is to build one model that disentangles the Y ~ X relationship into a general component, which summarises the general trajectory for all species, and a species-level component, which estimates how the relationship of Y ~ X is conditional on species identity. This is, essentially, a hierarchical model: we are estimating variation (1) across species, and (2) between species in one model.

Which comes first? The chicken or the egg?

Over the last decade, we’ve seen a burst in modeling capacity in ecology. We have big datasets, more powerful computers, a wide catalogue of statistical modeling tools and literature to back it up, and importantly – a ton of questions we want to answer.

These questions should, of course, always be based on our previous understanding of relationships we’ve studied between Y and X from the literature and our observations of nature. Our questions are supposed to shape the model we build – but, we have to admit that sometimes the model shapes the question first. Maybe our labs use certain models that we’ve adopted ourselves, and we find ways to apply this model to new datasets. Maybe our field loves using a certain approach (like community ecologists love a good PCA – no shame here, I love a PCA too), so we reach for it first. Maybe we think a newly published modeling approach is fancy and cool, and want to find a way to use it in our research. These are all fine, honestly. 

But we are wondering: Are our research questions sometimes limited by our model “comfort zone”, and if so, how do we push past this?

In other words, like the chicken and the egg: Which came first, the question or the model?

Are there questions that we aren’t yet asking because we don’t know how to answer them? Are some of these questions actually possible to answer now that we have an abundance of models, types and sizes of datasets, and powerful computers?

Join a community call!

If you’re interested in talking more about these models, please join our community calls during March and April 2025! We welcome anyone interested in GAMs, computational ecology, or eager to learn more about HGAMs to participate in the following sessions:

These community calls are intended to help us face this question. Each discussion will focus on the outstanding ecological questions that we could answer with HGAMs, highlighting a wide array of potential applications for specific types of ecological and evolutionary data. Join us in thinking about how we could use HGAMs to push ecological research forward! You do not need any background with hierarchical modeling or generalized additive models to join these discussions.

Each discussion will follow this structure:

ActivityDuration
Welcome & Scope10 mins
Individual reflection before the small group discussion5 mins
Small group discussion: What question do you usually ask? What question would you like to ask next?What model(s) do you use to answer your questions? Are some of your questions (or questions you’d like to ask) something you could ask with a hierarchical model, or hierarchical GAM?15 mins
Whole group discussion (discuss the small group findings)25 min
Wrap up & next steps5 mins
Close

What is this all for?

Our intention is to collaboratively write a Perspective paper to highlight the outstanding questions in ecology that could be explored with hierarchical GAMs. 

Please fill this form to let us know how you would like to participate (or not) in the next steps. 

All participants of this call who contributed to the notes and discussions will be credited in the acknowledgements of the paper unless otherwise communicated to us. 

Authored by Camille Lévesque and Katherine Hébert

HGAMs working group community calls

We invite you to join our series of discussions on the applications of Hierarchical Generalized Additive Models (HGAMs) in ecology. These discussions are part of a broader initiative led by BIOS2’s HGAMs working group, aimed at promoting the understanding and application of these models. 

We welcome anyone interested in GAMs, computational ecology, or eager to learn more about HGAMs to participate in the following sessions:

Each discussion will focus on the outstanding ecological questions that we could answer with HGAMs, highlighting a wide array of potential applications for specific types of ecological and evolutionary data. Join us in thinking about how we could use HGAMs to push ecological research forward!

Student-Led research group call 2024/25

The concept and idea of a “working group” varies immensely in academia, but in the industry, it has a very clear definition: it is a group of people working on a very specific problem within a timeframe. Although the time to digest and rethink deeply complex ecological problems is instrumental for scientific progress, very often scientists are called to act, to propose a specific solution for a specific problem (even more in the rapidly changing world we’re living in).

The Student-Led Working group aims to promote leadership opportunities for BIOS2 fellows while fostering diversity, inclusion and accessibility in research environments. This call simulates a project development pipeline, with four phases: initiation, planning, execution and closing. Proponents of a working group would submit a project where the organization and leadership are well-defined in the project, as well as expectations, budget and timelines. BIOS2 fellows are invited to submit a project proposal with a schedule for up to seven months, including at least the execution and closing phases of a research project.

A budget of up to CAD$ 5000 is available for travel, accommodation, material and publication fees. In-person event coordination and/or mentoring for organizers can be provided by the BIOS2 coordination team.

Project submission

Each project would clearly define a roadmap including the four phases of project development cited above.

The project should not last more than one academic year and can be very short to include only the final phase of the working group. For example, if a student have already gone through the ideation phase, but needs help planning and executing their projects, they should include the ideation phase in the project and describe the outcomes that would be the input of the planning phase.

Additionally, the group members should be defined by the proponent. A very important aspect of a working group is the choice of the people who will work together. It requires planning and leadership skills to identify the people with the technical baggage, interests and availability to work on a project. It’s up to the proponent how they will find the right people. If needed, the applicant group can open a call for applications to invite more participants, and BIOS2 can help publicizing this call, but the final selection should be made entirely by the proposal’s leadership group. BIOS2 will consider the diversity of the group and justifications for the group composition when deciding if the project will be funded.

The project may or may not include in-person meetings, but should include at least one sprint period. Sprints are short, time-boxed periods when a team works to complete a very well-defined set amount of work. The amount of work defined by the leaders must be compatible with the length of the sprint: the shorter the period, the smaller the number of tasks to be completed. An imbalance in this ratio will certainly lead to burnout and frustration. The sustainability and inclusivity of in-person meetings must be considered during the selection process.

The application package will consist of a written project (that can be supported by other media, such as videos, audio, images, websites, etc.), a lean canvas proposal, a budget spreadsheet and a timeline of ideation (when applicable), planning, execution and closing phases. This timeline must comprise any period between December 1st, 2024 and August 31st, 2025, including the closing phase.

The written project and supporting media should expand on the bullet points added to the lean canvas, and there, proponents must define:

  • One clear problem to be solved and why it’s important to solve it. Applicants should include their problem in a good theoretical context that can be understood by specialists and non-specialists.
  • Description of the human resources of the group, i.e., names and contributions of each participant. Applicants must not attach any CVs – instead, they should explain why they decided to invite these people to help them solve the problem defined before.
  • Expected outcomes: explicit what and how many products they expect to deliver by the end of the timeline, why they are important, and how they can be used as a measure of success.
  • The budget document must be filled in in detail. Applicants should include itemized expenses instead of broad categories (i.e., instead of “support for participants”, they should try to describe which kind of support they are talking about – child care? transportation? translation? masks?).
  • Applicants must also include a description of regular activities, depending on the number of phases they include in the project (frequency of virtual and in-person meetings, visits, training, etc.).

The templates for the lean canvas and budget are available below:

Eligibility

Applicants must be currently registered in a study program in ecology, environmental science, evolution or a related discipline in a Canadian university and be a BIOS2 fellow. The other members of the group don’t need to fulfill these requirements, but only students can receive funds from BIOS2. Non-students in the group should be funded by other sources.

Selection process and deadlines

Applications must be submitted by November 8th, 2024. The selection committee will be composed of BIOS2 and QCBS staff members and alumni. Results are to be expected in the last two weeks of November, 2024.

Application packages should be submitted as one single zip file through this form: https://tally.so/r/31vypW

Several calls for applications are now open!

The 2023 season of call for applications of BIOS2 is open! There are several opportunities available: undergraduate fellowship, postdoctoral fellowship for persistence and graduate student fellowship. Applications for the undergraduate and postdoctoral fellowships will be open until the positions are filled, and the graduate student fellowship applications close on August 18, 2023.

Call 2022 to participate in the BIOS² program

This is the 2022 Call for BIOS² Fellows. This call is for current BIOS2 Fellows to renew their membership and/or funding and for graduate students to join the BIOS2 program.

Applications are due by August 10, 2022.

Program Description

The Computational Biodiversity Science and Services BIOS2 training program is a NSERC-CREATE program. Fellows learn computational and quantitative skills from some of Canada’s best biodiversity scientists and apply skills to solve real-world problems through internships and working groups. BIOS² aims at widening opportunities and skill sets among students and postdoctoral fellows and increasing recruitment in biodiversity science in the Canadian job market.

BIOS2 training consists of instruction (courses, short modules, and summer schools), working groups and internships. Fellows will use their internships and working groups to solve real-world problems related to biodiversity and to get valuable professional experience outside of academia.

Call for proposals: Open projects in biodiversity science and Non-NSE domains

The BIOS2 Research Funding Program supports open projects in biodiversity science and non-NSE domains proposed by academic members

The BIOS² training program provides a framework to foster collaborative, multidisciplinary and cross-sectoral research. It focuses on training the next generation of highly qualified people in quantitative ecology as well as future actors and decision makers in the biodiversity science sector. The main objectives of the program are to widen opportunities and skill sets among students and increase their recruitment in Biodiversity science in the Canadian job market. 

We are currently offering financial support for projects in biodiversity science, and in non-Natural Sciences and Engineering domains. The aim of this call is to fund projects, not students. Once the project is awarded, students are selected by the academic member(s) leading the project.