Francisco Rodriguez-Sanchez

@frodsan.bsky.social

3470 Followers 173 Following

Computational Ecologist. Associate Professor @unisevilla.us.es. Ecology, biogeography, statistics, rstats, GIS, science. https://frodriguezsanchez.net

ir makes the runtime specification part of the source file. That means a script can say, directly: which packages it needs which R should run it, when that needs to be explicit whether user libraries should be visible whether CRAN packages should be resolved as of a specific date This should help you rerun the script reliably at a later date. Keeping the metadata in the file makes it less likely to be lost or fall out of sync with the code.
Profiting from pilot studies: Analysing mortality using Bayesian models with informative priors Pilot studies are often used to help design ecological studies. Ideally the pilot data are incorporated into the full-scale study data, but if the pilot study's results indicate a need for major changes to experimental design, then pooling pilot and full-scale study data is difficult. The default position is to disregard the preliminary data. But ignoring pilot study data after a more comprehensive study has been completed forgoes statistical power or costs more by sampling additional data equivalent to the pilot study's sample size. With Bayesian methods, pilot study data can be used as an informative prior for a model built from the full-scale study dataset. We demonstrate a Bayesian method for recovering information from otherwise unusable pilot study data with a case study on eucalypt seedling mortality. A pilot study of eucalypt tree seedling mortality was conducted in southeastern Australia in 2005. A larger study with a modified design was conducted the following year. The two datasets differed substantially, so they could not easily be combined. Posterior estimates from pilot dataset model parameters were used to inform a model for the second larger dataset. Model checking indicated that incorporating prior information maintained the predictive capacity of the model with respect to the training data. Importantly, adding prior information improved model accuracy in predicting a validation dataset. Adding prior information increased the precision and the effective sample size for estimating the average mortality rate. We recommend that practitioners move away from the default position of discarding pilot study data when they are incompatible with the form of their full-scale studies. More generally, we recommend that ecologists should use informative priors more frequently to reap the benefits of the additional data.
Rule 8: Commit to tidyverse or base R In R, you can help new learners build a strong foundation by committing to either the Tidyverse ecosystem or base R from the outset. Switching between related languages or language “dialects” can erode learner confidence and cause confusion [21,34]. For example, learners often struggle with when to extract a column of a dataset as a vector using dplyr::pull(). Mixing in base $ notation further muddies the waters. A new programmer can accomplish their end goal using either readr::read_csv (Tidyverse) or read.csv() (base R) and does not need to know both when starting out. The choice of R ecosystem depends both on your experience and comfort level, as well as the goals of your student audience. For example, our audience typically consists of researchers and practitioners in the public health sphere who have no prior R experience. These learners are more interested in data cleaning, analysis, and visualization. The Tidyverse works well for these goals and is valued by instructors for its readability, consistency, and user community [34]. However, if your learners are more interested in software development, dependency minimization, optimization, or mathematical applications, base R might be more appropriate. Some researchers have found base R easier for learners [35], while others found no difference [36].
This image shows a workflow for placeholder-style reproducible manuscripts. In this case, data and analysis code are combined to produce analysis outputs. Separately, a manuscript text is written and collaborated on, containing placeholders for where analysis values will go. A script knits these two together, inserting analysis values where the placeholders are, producing the final populated manuscript.
Jérôme Lecoq, a neuroscientist at the Allen Institute, used Claude Science to build a multi-agent “computational review template” comprising about 20 custom skills geared towards writing long-form reviews. The sub-agents read through thousands of papers, pulling the central claim and the key quantitative finding, and storing them in an evidence state database. Then the pipeline constructs a narrative arc, writing the review section by section and delegating each to its own specialized sub-agent. Within each section, dedicated agents generate quantitative cross-study figures directly from the evidence database. A key component of the workflow, enabled by Claude Science, is the use of actor-critic pairs: one agent creates content while a separate reviewer agent evaluates it for accuracy and citation fidelity. Before Claude Science, it could take Lecoq’s team as many as two years to write such a review. He now has about 10 reviews, many more than 100 pages, with citations that were checked over by reviewer agents. The team is now working with domain experts to further refine the AI-based critic agents.
For the past two decades, the message to young ecologists has been consistent: learn to code, learn statistics, learn machine learning, learn ‘modelling’ (GIS, differential equations, species distribution modelling, etc.). That message was correct and remains so. But it is incomplete. The researchers who will hold a durable comparative advantage in the decades ahead may be those who invested in the patient, expensive, logistically difficult work of building field programs, maintaining experimental infrastructure and cultivating the expertise required to collect data that no one else can collect.
One or two sentences providing a basic introduction to the field, comprehensible to a scientist in any discipline. Two to three sentences of more detailed background, comprehensible to scientists in related disciplines. One sentence clearly stating the general problem being addressed by this particular study. One sentence summarizing the main result (with the words “here we show” or their equivalent). Two or three sentences explaining what the main result reveals in direct comparison to what was thought to be the case previously, or how the main result adds to previous knowledge. One or two sentences to put the results into a more general context. Two or three sentences to provide a broader perspective, readily comprehensible to a scientist in any discipline, may be included in the first paragraph if the editor considers that the accessibility of the paper is significantly enhanced by their inclusion.
Summary of a paper’s structural elements at three spatial scales: Across sections, across paragraphs, and within paragraphs.
A structured workflow for drafting a scientific paper in 15 Steps. Use this figure as a practical map: work through it sequentially to build a manuscript from scratch, or revisit individual steps to diagnose problems and guide revision.
Querying data from large cloud databases with R and DuckDB A few years ago I became fascinated with cloud-friendly data formats that allow us to query large online databases without having to download the entire thing. Here is a post from 2021 on accessing spatial data from online cloud-optimised geotiffs using R, terra and gdal. In this post I will show how to access, manage, and retrieve data from large tabular databases hosted in the cloud, without having to download them first. As you will see, thanks to recent efforts from many parties, the process is easy and very fast.
Example of querying cloud database with duckplyr in R
High VIF values indicate that coefficient estimates may be unstable and have inflated standard errors. However, removing predictors with high VIF values is generally not recommended as a blanket solution (Vanhove 2021; Morrissey and Ruxton 2018; Gregorich et al. 2021). Multicollinearity is primarily a concern for interpretation of individual coefficients, not for the model’s overall predictive performance or for drawing inferences about the combined effects of correlated predictors. Consider these points when dealing with multicollinearity: If your goal is prediction, multicollinearity is typically not a problem. The model can still make accurate predictions even when predictors are highly correlated (Feng et al. 2019; Graham 2003). If your goal is to interpret individual coefficients, high VIF values signal that you should be cautious. The coefficients represent the effect of each predictor while holding all others constant, which may not be meaningful when predictors are strongly related. In such cases, consider: Interpreting coefficients jointly rather than individually Acknowledging the uncertainty in individual coefficient estimates Considering whether your research question truly requires separating the effects of correlated predictors For interaction terms, high VIF values are expected and often unavoidable. This is sometimes called “inessential ill-conditioning” (Francoeur 2013). Centering the component variables can sometimes help reduce VIF values for interactions (Kim and Jung 2024). Consider the substantive context: Sometimes, multicollinearity reflects important aspects of your data or research question. Removing variables to reduce VIF may actually harm your analysis by omitting important confounders or by changing the interpretation of remaining coefficients (Gregorich et al. 2021).
major stumbling blocks for AI include routinely undervaluing work awarded top marks by humans, or overvaluing essays ranked among the lowest. Unlike human examiners, all the AI systems were “oversensitive to linguistic features”: giving out higher marks based on essay length, vocabulary range and sentence complexity, which are often unrelated to academic standards. In the latest report, researchers suggest that AI could be valuable for aspects of student assessment such as error detection and consistency checks – a “second pair of eyes” – as well as triaging feedback for students.
Note Users should keep in mind the hazards that a “thoughtless approach” of evaluating all possible models poses. Although this procedure is in certain cases useful and justified, it may result in selecting a spurious “best” model, due to the model selection bias. “Let the computer find out” is a poor strategy and usually reflects the fact that the researcher did not bother to think clearly about the problem of interest and its scientific setting (Burnham and Anderson, 2002).
What hurts, what helps: mental health in supervisor - ECR relations. Perceived impact of supervisor behavior on mental health and frequency of particular behaviors. A) Overall perceived impact of supervisors’ behavior on respondents’ mental health. B) Percentage of respondents reporting specific supervisor behaviors with a negative impact. C) Percentage of respondents reporting specific supervisor behaviors with a positive impact. Respondents selected their top three behaviors each category (positive or negative).
Multi-response phylogenetic mixed models: concepts and application The scale and resolution of trait databases and molecular phylogenies is increasing rapidly. These resources permit many open questions in comparative biology to be addressed with the right statistical tools. Multi-response (MR) phylogenetic mixed models (PMMs) offer great potential for multivariate analyses of trait evolution. While flexible and powerful, these methods are not often employed by researchers in ecology and evolution, reflecting a specialised and technical literature that creates barriers to usage for many biologists. Here we present a practical and accessible guide to MR-PMMs. We begin with a review of single-response (SR) PMMs to introduce key concepts and outline the limitations of this approach for characterising patterns of trait coevolution. We emphasise MR-PMMs as a preferable approach for analyses involving multiple species traits, due to the explicit decomposition of trait covariances. We discuss multilevel models, multivariate models of evolution, and extensions to non-Gaussian response traits. We highlight techniques for causal inference using graphical models, as well as advanced topics including prior specification and latent factor models. Using simulated data and visual examples, we discuss interpretation, prediction, and model validation. We implement many of the techniques discussed in example analyses of plant functional traits to demonstrate the general utility of MR-PMMs in handling complex real-world data sets. Finally, we discuss the emerging synthesis of comparative techniques made possible by MR-PMMs, highlight strengths and weaknesses, and offer practical recommendations to analysts. To complement this material, we provide online tutorials including side-by-side model implementations in two popular R packages, MCMCglmm and brms.
In each city, two observers—one woman and one man—collected data at all sites (in total, four women and four men, all expert ornithologists, were involved in the study). The observer pairs were of similar height and wore similarly coloured clothes during the sampling trials (white, grey or black). We compared sex-specific data on the height and body mass of observers by using the Wilcoxon signed rank test for paired samples. The mean difference in observers' height by sex was 5.5 cm, while the mean difference in observers' body mass was 7.3 kg; however, neither of which was statistically different between women and men observers
In this paper, we discuss an extension to two popular approaches to modeling complex structures in ecological data: the generalized additive model (GAM) and the hierarchical model (HGLM). The hierarchical GAM (HGAM), allows modeling of nonlinear functional relationships between covariates and outcomes where the shape of the function itself varies between different grouping levels. We describe the theoretical connection between HGAMs, HGLMs, and GAMs, explain how to model different assumptions about the degree of intergroup variability in functional response, and show how HGAMs can be readily fitted using existing GAM software, the mgcv package in R. We also discuss computational and statistical issues with fitting these models, and demonstrate how to fit HGAMs on example data. All code and data used to generate this paper are available at: github.com/eric-pedersen/mixed-effect-gams.
Foundational training in ecology is overwhelmingly focused on competition and predation, relegating mutualism to the periphery. This decades-long pedagogical bias restricts our theoretical frameworks to negative interactions, creating a deeply-rooted theory-empiricism disconnect in mutualism research. Lacking a dynamic theoretical framing, empiricists frequently collect data that cannot be integrated into population models, limiting our capacity to predict ecosystem responses to environmental change. To transform mutualism into a robust component of general ecology, this forum piece distills a recent theoretical synthesis into an accessible framework that generates testable predictions on par with antagonistic interactions. After tracing the historical recognition of positive interactions, I distill the core population dynamics of mutualism to reveal its universal signatures: stable coexistence driven by benefit saturation, and destabilizing thresholds at low and extremely high densities. I then outline a roadmap for testing these theoretical predictions empirically. Ultimately, integrating this framework into the foundational ecological canon will break the pervasive pedagogy-research cycle, establishing the iterative dialogue between models and data necessary to build a general theory of positive interactions.
A transparent universal credit system to incentivize peer review Science relies on peer review, but it’s getting harder to find people willing to do it (1–3). Between 2013 and 2017, the average number of invitations for each review increased by 26%, and the number of review invitations grew twice as fast as the number of articles published (4). Three-quarters of journal editors say the hardest part of their job is finding willing reviewers (4). Submission rates to individual journals are also increasing, placing further strain on the peer-review system (5).
ColorBrewer diverging palettes
Características del contrato: se requiere Grado en Biología, Ciencias Ambientales o Ingeniería Informática, además de certificado de inglés nivel B1. Se trabajará en: Recopilación de datos sobre características de frutos y semillas de plantas de todo el mundo. Uso de herramientas informáticas para la extracción masiva de datos de fuentes no estructuradas (texto, imágenes). Integración de los datos en una base de datos global utilizando R y herramientas de control de versiones (git, GitHub Actions).
In this article, we present the case for Generalized Linear Latent Variable Models (GLLVMs) as a go-to choice of statistical method for any community ecologist wanting to tackle a range of present-day ecological research questions. GLLVMs bring tools and capabilities from classic (mixed-effects) regression models to multivariate community analysis, providing a number of novel ways to tailor models specifically to one’s study questions and data properties not available when using non-model-based multivariate methods. In order to facilitate further adoption of these methods by community ecologists, we provide 1) a practitioner-focused and practical overview of the advantages the GLLVM framework brings to the table when addressing different core ecological questions, 2) a number of concrete suggestions for how GLLVMs best can be incorporated into the analytical workflow of community ecologists, and 3) two illustrative worked examples of this workflow in action on real-world data.
You usually don’t want to test for fit to a Normal distribution. If you do, you usually don’t care primarily about the alternatives to which the Shapiro-Wilk test is sensitive. Unless the conclusion is very clear, pre-testing is going to completely munt your subsequent tests. And if the conclusion is clear then you can look at the qq-plot; you don’t need to get maths to look at it for you, and you might learn something else about the data.
A Shapiro-Wilk test of the response variable concludes very significant deviation of Normality. But residuals of linear model consistent with Normal distribution.
How AI assistance impacts coding speed and skill formation
CRAN DOI-generated citation of ggplot2 package: Wickham, H., Chang, W., Henry, L., Pedersen, T. L., Takahashi, K., Wilke, C., Woo, K., Yutani, H., Dunnington, D., Brand, T. van den, Posit, & Pbc. (2025). ggplot2: Create elegant data visualisations using the grammar of graphics (Versión 4.0.1) [Software]. https://cran.r-project.org/web/packages/ggplot2/index.html
Examples of how to use the new futurize package
Almond flower
Should PhD programmes require candidates to have at least one paper published in order to graduate? Many PhD programmes require PhD candidates to have one (or more) papers accepted or published in a scientific journal in order to be able to graduate and complete their doctorate. This is of course a well-intentioned idea and may sound good at first, as PhD candidates would gain experience and start learning how to navigate the publication process in scientific journals. Having a publication out the PhD may also help them secure their next position. The publication requirement would also encourage that research results are published and not lost forever, particularly if the new PhD moves on after completing their doctorate. Finally, having a paper accepted at a scientific journal usually means that at least three people (editors and reviewers) have evaluated the research and the editor(s) have considered it good enough for publication, so it could be interpreted as an external validation and signal of quality. Undesired consequences of requiring publication in PhD programmes So far so good. But on second thoughts one can foresee how this publication requirement can create perverse incentives, bring many problems for early career researchers and science in general, and basically do more harm than good. Particularly considering that many PhD programmes last 3-4 years (at least in Europe) and the publication process can take well more than a year from submission to acceptance (more on this below). Thus, a few counter-arguments that come to mind: [continues in the linked website]
Geography of masting creates greater synchrony in seed scarcity than in seed abundance Interannually highly variable and synchronized production of large seed crops by perennial plants, called masting, drives resource pulses and famines with cascading effects on food webs. While the spatial scale of masting synchrony is well documented, it remains unclear how synchrony differs between years of seed abundance and failure, and how such dynamics extend across species and space. These gaps are important to resolve, as they determine the magnitude and spatial extent of masting effects on food webs. Using a 36-y dataset from 431 sites spanning seven dominant tree species in temperate Europe, we provide evidence that seed failures are more spatially synchronized than mast peaks, indicating that regional coherence in seed production is structured primarily by reproductive failure. Among-species synchrony was localized. This suggests that in temperate forests, mobile seed consumers are unlikely to experience coordinated starvation–satiation cycles, in contrast to highly synchronous tropical dipterocarp systems. From an applied perspective, failure years affect seed availability over broad regions, limiting sourcing options for afforestation and restoration, and underscoring the value of spatially explicit masting forecasting. Because mast peaks and failures differ fundamentally in their food web consequences, our findings highlight the need to better understand and anticipate the ecological impacts of synchronized seed scarcity.
Mapas y rutas de sombra: cómo la ciencia de datos optimiza nuestros pasos bajo el sol
Mapas y rutas de sombra: cómo la ciencia de datos optimiza nuestros pasos bajo el sol
Contenidos del curso
The dataset contains nine 0s and one 1. One approach is to nudge our 0s towards .01 and our 1s to .99, or apply a special formula (Smithson & Verkuilen, 2006) so values fall within the [0, 1] interval. We implore readers not to engage in this practice. Kubinec (2022) showed that this practice can result in serious distortion of the outcome as the sample size grows larger, resulting in ever smaller values that are “nudged”. Because the Beta distribution is a non-linear model of the outcome, values that are very close to the boundary, such as 0.00001 or 0.99999, will be highly influential outliers.
Collaborative development of R packages for ecology https://ecologyr.github.io/workshop/
Recognizing and responding to a replication crisis: Lessons from Psychology The last 15 years have been a turbulent period for the field of psychology. Failures to replicate, fraud, and evidence of questionable research practices have all contributed to a crisis of credibility. What has the field done in response to this crisis, and what lessons can be learned from psychology’s crisis? We often hear the self-correcting mechanisms in science invoked as a reason to trust science, but it is not clear what these mechanisms are, or how they could have prevented psychology’s crisis. Some quality control mechanisms, such as peer review for journals, or vetting for textbooks or for public dissemination, did not to provide much of a safeguard against invalid claims in psychology. Instead, I argue that we should look for visible signs of a scientific community’s commitment to self-correction. These signs include transparency in the research and peer review process, investment in error detection and quality control, and an emphasis on calibration rather than popularization. To regain or bolster trust, scientific fields should invest in these hallmarks of credibility.
Book cover of the Chicago guide to college science teaching
A screenshot of a statistics tutorial made with R, quarto and webR
library("pins") # specify folder in Google Drive to store objects board <- board_gdrive("https://drive.google.com/drive/folders/FOLDER_ID") # upload file board |> pin_write(big_dataset, "big_dataset.csv", type = "csv") # retrieve file (e.g. in another computer) board |> pin_read("big_dataset.csv")
Now possible to scan and cite dependencies from a package DESCRIPTION file (#62). Use pkgs = c("Depends", "Imports", "Suggests", "LinkingTo") or combinations of them to choose which dependencies to cite. Now possible to scan and cite dependencies from a single R script, ‘Rmarkdown’ or ‘Quarto’ document #65. Just provide the path to the file to the pkgs argument. Now possible to customise the language of citation paragraphs (#55), thanks to new arguments to cite_packages: text.start, text.pks and text.RStudio. New logical argument skip.missing allows the user to skip missing packages (those used somewhere in the project but not currently installed) #54. Slow down requests when querying dependencies for a large number of packages #61.
Joint species-trait distribution modeling: The role of intraspecific trait variation in community assembly The links between intraspecific trait variation and community assembly remain little studied, partially due to the lack of statistical methods to jointly model intraspecific trait variation and species abundances at the community level. Here, we extend the joint species distribution modeling (JSDM) framework into the joint species-trait distribution modeling (JSTDM) framework to explicitly link species abundances to phenotypic variation in traits for multiple species simultaneously. Using a case study of 65 tundra plant species abundances and 3 key functional traits measured across 325 sites, we show how the JSTDM approach (1) estimates the statistical associations among species abundances, species-level traits, and site-level traits, relative to environmental variation; (2) improves predictions on trait variation by using information on species abundances; and (3) generates hypotheses about trait-driven community assembly mechanisms. The JSTDM methodology presented in this study allows assessing the interplay between species abundances and traits at the community level, providing the much needed modeling tools to quantify the role of phenotypic trait variation in eco-evolutionary community assembly.
In the simplest scenario, there is high bias in the interaction parameter (used for co-occurrence inference) with less than 100 sites at high detection, and 400–1000 sites at low detection, depending on interaction strength. Strong co-occurrence is detected consistently above 200 sites with high detection probabilities, but weak co-occurrence is never consistently detected even with 2980 sites. We demonstrate that the mean predictive ability of the co-occurrence model is less affected by sample size
A table of packages generated with grateful::cite_packages() where some packages have been omitted using the new argument `skip.missing = TRUE`
A table of {grateful} package dependencies, created with grateful::cite_packages()
A package citation paragraph in Spanish created with grateful::cite_packages()
grateful: Facilitate citation of R packages The goal of grateful is to make it very easy to cite R and the R packages used in any analyses, so that package authors receive their deserved credit. By calling a single function, grateful will scan the project for R packages used and generate a BibTeX file containing all citations for those packages. grateful can then generate a new document with citations in the desired output format (Word, PDF, LaTeX, HTML, Markdown). These references can be formatted for a specific journal, so that we can just paste them directly into our manuscript or report. Alternatively, we can use grateful directly within an Rmarkdown or Quarto document. In this case, a paragraph containing in-text citations of all used R packages will (optionally) be inserted into the Rmarkdown/Quarto document, and these packages will be included in the reference list when rendering.
Understanding functional diversity, the range and variability of species' roles and actions within their communities, is key to predicting and preserving the functions that sustain both nature and human well-being. In this paper, we provide a comprehensive review of the literature on functional diversity measurement. We begin by consolidating essential criteria that effective measures of functional diversity should meet. We then evaluate fifteen widely used functional diversity metrics against these criteria and assess their performance across six synthetic ecosystem scenarios where optimal behavior is known. Surprisingly, our analysis reveals that none of the widely used metrics fully satisfy all the established requirements, and all fail in at least one ecosystem scenario. In particular, we find that almost all metrics flagrantly violate set monotonicity and distance monotonicity, requirements that adding a novel species should increase diversity, and that the magnitude of that increase should grow with trait dissimilarity. We also find that metrics fail to decline when rare, functionally extreme species are lost, and even increase when a perfectly redundant species is added. These critical flaws leave them blind to the very biodiversity loss that functional diversity measures are intended to detect. Our findings underscore the urgent need to develop a new generation of functional diversity metrics that more accurately reflect ecological realities.
Interactive function map of an R package, generated with https://uptake.github.io/pkgnet
Function map for the 'grateful' package: https://pakillo.github.io/grateful/
There is currently a dichotomy in the modelling of Grinnellian and Eltonian niches. Despite similar underlying data, Grinnellian niches are modelled with species-distribution models (SDMs), whereas Eltonian niches are modelled with ecological-network analysis, mainly because the sparsity of species-interaction data prevents the application of SDMs to Eltonian-niche modelling. Here, we propose to adapt recently developed joint species distribution models (JSDMs) to data on ecological networks, functional traits, and phylogenies to model species' Eltonian niches. JSDMs overcome sparsity and improve predictions for individual species by considering non-independent relationships among co-occurring species; this unique ability makes them particularly suited for sparse datasets such as ecological networks. Our Eltonian JSDMs reveal strong relationships between species' Eltonian niches and their functional traits and phylogeny. Moreover, we demonstrate that JSDMs can accurately predict the interactions of species for which no empirical interaction data are available, based solely on their functional traits. This facilitates prediction of new interactions in communities with altered composition, for example, following climate-change induced local extinctions or species introductions. The high interpretability of Eltonian JSDMs will provide unique insights into mechanisms underlying species interactions and the potential impacts of environmental changes and invasive species on species interactions in ecological communities.
df <- data.frame( web = c("https://example.com/", "https://example.com/"), value = c(NA, NA) ) for (i in 1:nrow(df)) { if (is.na(df$value[i])) { browseURL(df$web[i]) df$value[i] <- menu(c("option A", "option B", "option C")) } }
df <- data.frame( web = c("https://example.com/", "https://example.com/"), value = c(NA, NA) ) for (i in 1:nrow(df)) { browseURL(df$web[i]) df$value[i] <- readline(prompt = "Please enter value: ") }