Round Table Discussions

ROUND TABLE DISCUSSION 1
Date: Tuesday, 28 June 2022
Time: 08h00 – 09h00 SAST (GMT +2)
Venue: On-line only

Defining power in Bayesian statistics
Dr Diego Barneche1, Dr Rebecca Fisher1, Dr Murray Logan1, Prof James McGree2, Dr David Warne2
1 Australian Institute Of Marine Science, Perth, Western Australia, Australia
2 Queensland University of Technology, Brisbane, Queensland, Australia

Power analysis has been a cornerstone of experimental design, so much so that it has become a mandatory requirement to justify, for example, usage of animal models or allocation of resources to experimental work. In some countries, regulatory agencies require environment risk assessment experiments to demonstrate a minimum power to detect a defined effect size. Statistical power is a well-defined binary hypothesis test in frequentist statistics: the probability of a test correctly rejecting the null hypothesis (i.e. when the alternative is true). On the other hand, in Bayesian statistics the notion of statistical significance is often rejected, with “significance” being largely inferred a posteriori based on the relevance of the findings to a particular field, conditional on parameter posterior probabilities. In the absence of a (still subjective) consensus for an acceptance rate—such as that of a p-value in frequentist statistics—hypothesis testing becomes even more subjective in Bayesian statistics. Therefore, different methods have been proposed to calculate statistical power in Bayesian statistics. This round table discussion aims to encourage participants to share their approaches to calculating power in Bayesian statistics, and perhaps reach a consensus on what might be an appropriate definition and/or implementation.

ROUND TABLE DISCUSSION 2
Date: Tuesday, 28 June 2022
Time: 08h00 – 09h00 SAST (GMT +2)
Venue: On-line only

Should the statistical ecology community create a new, diamond open access journal?
Dr Frédéric Barraquand1
1 CNRS, Talence, France

Statistical ecology develops at a great pace while the publishing model is getting increasingly unsustainable, with big publishers dominating the market. While statistical ecology can be published in either statistics or ecology, there is a rationale for publishing at the interface. But such society-based journals are few and Methods in Ecology and Evolution now publishes articles from all methodological subfields, creating an intense pressure for space. MEE’s cascade journal, Ecology & Evolution, has more and more statistical ecology papers but this route favours the gold OA model which is unfair to scientists from less well-off countries, and renders statistical ecology less visible. There may therefore be a scientific niche for new society-based statistical ecology journal.

Moreover, we believe that it would be time to favour a fairer diamond open access publishing model, sponsored by the institutions, at no cost to either authors or readers. We present multiple options for diamond/fair open access, and ask whether the ISEC community should focus on creating a specialised diamond OA statistical ecology journal (following successes in mathematics), or simply aggregate papers published in more generalist fair open access initiatives – or both. A dedicated journal would allow to publish high-quality papers a little more statistical in nature than at MEE, as well as the increasingly many works refining statistical methods to hone in on ecological questions. On the other hand, a new dedicated journal would require time and support from the statistical ecology community

ROUND TABLE DISCUSSION 3
Date: 28 June 2022
Time: 08h00 – 09h00 SAST (GMT +2)
Venue: Venue 5 & On-line

Integrated population models: brainstorming towards the organisation of a “best practices” workshop
Dr Chloé Nater1, Dr Olivier Gimenez2, Dr Marc Kéry3, Dr Michael Schaub3, Dr Elise Zipkin4
1 Norwegian Institute For Nature Research, Trondheim, Norway
2 CNRS, Montpellier, France
3 Swiss Ornithological Institute, Sempach, Switzerland
4 Michigan State University, East Lansing, Michigan, United States

Integrated population models (IPMs) have rapidly gained popularity over the last years. The large potential of data integration, combined with the availability of accessible learning material and easy-to-use and powerful software, have led to a real boom in the number of applications. Recent IPMs can now include a wide variety of data types, model continuous traits, account for spatial structure, multi- and meta-population dynamics, species interactions, and much more. Concurrent with these developments, the IPM user base has also grown and diversified greatly. Looking at papers on and analysis code for IPMs, one cannot help but notice large variation in how and to what degree common challenges are addressed, as well as in the quality of documentation, accessibility, reproducibility, and re-use potential of models and workflows. New and experienced users alike could benefit greatly from having some general guidelines for navigating challenges and building models and workflows that can optimize their own project, as well as being useful beyond a single research paper.

Such guidelines for best practices should be developed within the community of IPM users and we aim to organize a workshop dedicated to this in 2023. We would like to use ISEC 2022 as an opportunity to meet with colleagues for discussing potential formats and practicalities for this workshop.

ROUND TABLE DISCUSSION 4
Date: 28 June 2022
Time: 08h00 – 09h00 SAST (GMT +2)
Venue: Venue 6 & On-line

Statistical Ecology in Africa
Prof Res Altwegg1, Dr Natasha Karenyi1, Dr David Maphisa1, Mr Mzabalazo Ngwenya1
1 Seec-UCT, Cape Town, South Africa

Africa has a rich biodiversity that underpins human livelihoods. This biodiversity is under pressure from various drivers and needs to be managed ever more carefully. Fortunately, data on how ecosystems respond to pressures are becoming more easily available across the continent. Likewise, the statistical tools to analyse these data are becoming more powerful and sophisticated. This is the domain of statistical ecology. While statistical ecology is growing quickly in Africa, we feel that capacity in this area lags behind the needs. In this round-table discussion, we will discuss ways for strengthening statistical ecology in Africa. How can we better support each other in our research and training? What are the challenges and opportunities? What are the funding opportunities we could tap into? We would like to bring together people who are interested in collaborating on statistical ecology across Africa and discuss setting up a continent-wide network in this field.

ROUND TABLE DISCUSSION 5
Date: 28 June 2022
Time: 14h00 – 15h00
Venue: Venue 4 & On-line

Steps to increase diversity and accessibility of ISEC and ecological statistics.
Dr Rocío Joo2, Dr Gordana Popovic1
1 UNSW Sydney, Sydney, NSW, Australia
2 Global Fishing Watch, Mexico/USA

There have been efforts by ISEC in the past few years to increase diversity though initiatives such as the bursary program. The move last year to an online conference also increased diversity, with many more people from developing countries attending than in previous years. However, there still is a large under representation from middle and lower income countries. Conferences in general are almost never accessible to people with hearing and visual impairments, representing a barrier for participation. We will discuss ideas for concrete steps to work towards greater diversity and inclusion in our community and conferences, including increasing ethnic and regional diversity, gender diversity, and accessibility for people with vision and hearing impairments and other disabilities.

ROUND TABLE DISCUSSION 6
Date: 28 June 2022
Time: 14h00 – 15h00
Venue: Venue 5 & On-line

A balanced review of multimodel inference in ecology
Mr Bert van der Veen1, Prof. Robert Brian O’Hara1
1 Norwegian University Of Science And Technology, Trondheim, Norway

Multimodel approaches are widely applied in ecology. Various papers have come out critising (Cade 2015) or advocating for their application (Araújo and New 2007). Examples of multimodel approaches that are frequently applied for ecological inference include model-averaging (Burnham and Anderson 2004), but also ensemble prediction (e.g., in the case of species distribution modeling; Geary et al. 2020). Many studies that attempt to predict species distributions, in truth look to explain an ecological process (Araújo et al. 2019). Though multimodel approaches might work well for prediction (Dormann et al. 2018) it remains unclear if their application also improves inference (Banner and Higgs 2017). In this discussion, we would like to go over some of the principles and motivations for multimodel inference, and discuss its suitability for ecological inference. Some key topics of the discussion could include; 1) what are best practices for multimodel inference in ecology?, 2) in the presence of multiple best fit models would it be better to describe uncertainty in the process rather than model-average?, 3) would ecological modeling benefit from a less explorative approach?, 4) should a single modeling approach be chosen a-priori on theoretical grounds?

Round Table Discussion 7
Date: 28 June 2022
Time: 14h00 – 15h00
Venue: Venue 6 & On-line

How to teach statistics to ecologists?
Tiago A. Marques, David Borchers, Chris Sutherland, Ali Johnson, Olivier Gimenez
1. University of St Andrews, UK
2. CNRS, France

The ISEC community is responsible for a considerable amount of the development of new methods that can be used for the analysis of ecological data. Therefore, we are as a community responsible for how the field evolves. Nonetheless, we feel that not enough consideration is given to what quantitative teaching should ecologists be subjected to maximize the effectiveness of the work they do, which at some level might be the uptake of newly proposed methods. This round table is a follow up to the same topic discussion held, and the questionnaire conducted, at the 2020 vISEC in Sydney, where we gathered the views from the ISEC community. We will present some of the results from that exercise to revive a discussion that seems to be much needed. We aim to, if a consensus can be reached, create outputs that might facilitate statistical training for ecologists. With that objective in mind we would like to promote discussion covering both what kind of methods ecologists should be exposed to but also how should the contents be delivered to them.