|
15h00 - 15h15
|
Welcome: Celebrating the Association's 10th anniversary
- Vera Pawlowsky-Glahn,
- Juan José Egozcue,
- Antonella Buccianti, and
- Dorothea Dumuid
Chair: Germà Coenders
|
|
15h15 - 15h50
|
- Speaker: Maria Isabel Orego, Universitat Politècnica de Catalunya, Spain
- Title: We must teach compositional data analysis to undergrads: On this hill I will die
- Abstract: Compositional Data Analysis (CoDA) has been around for decades, with its foundations rooted in the work of Aitchison in the 1980s, though its ideas can be traced back even further. Despite its growing use since the 2000s, CoDA is still not widely known, even among statisticians and practitioners. One reason for this is that many people only discover CoDA late in their careers when they encounter research problems that require a compositional perspective. And there is a wide variety of problems that require CoDa techniques. Although the first applications of CoDA mainly dealt with geoscience problems recent applications of the methodology span areas as diverse as microbiome tourism sequencing studies, archaeology, genetics, engineering, urbanism, water and sanitation, economy, energy, and planetology. In this presentation, I argue that CoDA should be introduced earlier in academic training, ideally at the undergraduate level. While not every student will use CoDA in their future work, those who do will benefit from recognizing compositional problems early and addressing them effectively. Based on my experience teaching Probability, Statistics, and related courses, I have introduced basic CoDA concepts to students with diverse backgrounds, from civil engineering undergraduates with strong algebra skills to master’s students with limited mathematical training. I’ll share some experiences where CoDA has been taught across different courses. Surprisingly, students often grasp these concepts more easily than expected, challenging the assumption that CoDA is too advanced for early learners. My findings suggest that integrating CoDA into general statistics education can provide students with a valuable tool for tackling real-world problems. This paper calls for a shift in how we teach statistics, making CoDA more accessible and widely understood across disciplines.
Chair: Gianna Monti
|
|
15h50 - 16h25
|
- Speaker: Christine Thomas-Agnan, Toulouse School of Economics, France
- Title: Scalar-on-composition and scalar-on-density compositional regression for assessing the impact of climate change on rice yield in Vietnam
- Abstract: The impact of temperature warming on many human activities is often analyzed with temperature temporal summaries, which can lead to a loss of valuable information. In econometrics, the impact of climate change on agricultural yield has sometimes been modeled using linear functional regression, where crop yield, a scalar response, is regressed on temperature density over time periods considered as a functional parameter, along with other covariates. We explore alternative models that summarize temperature data by their distribution over a given time period while preserving its distributional nature. These models are relevant for phenomenons which are insensitive to temporal order. Since classical addition and scalar multiplication are unsuitable for density functions, alternative operations and spaces are required. Moreover, compositional data analysis suggests that such covariates should undergo appropriate log-ratio transformations before inclusion in the model. We compare a discrete version with temperature histograms treated as compositional vectors and a smooth scalar-on-density regression using Bayes space representation for temperature densities. Using daily temperature extremes to model rice yield in Vietnam, we evaluate the advantages of each method. We also discuss parameters interpretation. Additionally, we propose modeling climate change scenarios via perturbations of the initial density, guided by a change direction curve derived from IPCC scenarios. The resulting rice yield impact (and its variance) are then quantified using a simple inner product between the density covariate parameter and the change direction curve. Our results indicate that while both approaches yield coherent findings, the smooth model outperforms the discrete one with an enhanced ability to accurately gauge the phenomenon’s scale.
Chair: Gianna Monti
|
|
16h25 - 17h00
|
- Speaker: Teresa Albuquerque, Polytechnic Institute of Castelo Branco, Portugal
- Title: Compositional Singularity in Geo-Exploration
- Abstract: The Valongo anticline is known for the existence of several Sb-Au and As-Au mineral deposits, many of which have been exploited since Roman times and are nowadays deactivated. For future exploration campaigns, the Baixo-Douro consortium (EDM-ECD-BRGM) conducted a campaign for stream sediment gathering, collecting 801 samples in this area, four per km2. The geochemical samples collected over approximately regular spatial grids have been analysed through spatial univariate singularity analysis to assess potential mineral deposits. However, the compositional nature of geochemical datasets brings the necessity of performing a compositional singularity assessment. The estimation of compositional singularities is accomplished through procedures that mirror the standard ones, except when dealing with raw concentrations that have not been log-ratio transformed, particularly for isometric or centre log-ratio transformed data. This survey introduces the concept of compositional singularity as a tool for designing future exploration strategies. Namely, the definition of preferential geological units and hotspots for the exploitation of Au-Sb. The spatial probability for significant spatial clusters in future exploration studies can be computed using the Simplicial Indicator Kriging. In this survey, Au and Sb computed singularities were used for illustration purposes. The threshold was selected using samples presenting high concentration-high singularity. Low-concentration-high singularity and high concentration-low singularity were used as a proxy for the expected error.
Chair: Gianna Monti
|
|
17h00 -
|
Conclusion
|
|
From 17h15
|
General Assembly of CoDa Association
|
