Searcharxiv⌕ Search

arXiv subjects

Robin J. Ryder

Publications and source records attributed to Robin J. Ryder.

14 recordsLinked to original sources
arXiv2025

Permutations accelerate Approximate Bayesian Computation

Approximate Bayesian Computation (ABC) methods have become essential tools for performing inference when likelihood functions are intractable or computationally prohibitive. However, their scalability remains a major challenge in hierarchical or high-dimensional models. In this paper, we introduce permABC, a new ABC framework designed for settings with both global and local parameters, where observations are grouped into exchangeable compartments. Building upon the Sequential Monte Carlo ABC (ABC-SMC) framework, permABC exploits the exchangeability of compartments through permutation-based matching, significantly improving computational efficiency. We then develop two further, complementary sequential strategies: Over Sampling, which facilitates early-stage acceptance by temporarily increasing the number of simulated compartments, and Under Matching, which relaxes the acceptance condition by matching only subsets of the data. These techniques allow for robust and scalable inference even in high-dimensional regimes. Through synthetic and real-world experiments -- including a hierarchical Susceptible-Infectious-Recover model of the early COVID-19 epidemic across 94 French departments -- we demonstrate the practical gains in accuracy and efficiency achieved by our approach.

stat.ME↗
arXiv2025

Phylogenetic latent space models for network data

Latent space models for network data characterize each node through a vector of latent features whose pairwise similarities define the edge probabilities among the pairs of nodes. Although this formulation has led to successful implementations, the overarching focus has been on directly inferring node embeddings through the latent features, rather than learning the generative process underlying these embeddings. This focus prevents borrowing information across the node features and limits the ability to infer higher-level architectures governing network formation. For example, routinely-studied networks often exhibit multiscale structures informing on nested modular hierarchies among nodes, which could be learned via tree-based representations of dependencies among the latent features. We pursue this direction by bridging latent variable representations of network data with concepts from phylogenetic inference to design a novel latent space model that explicitly characterizes the generative process of the node feature vectors through a branching Brownian motion, with branching structure parametrized by a tree. This tree constitutes the main object of interest and is learned under a Bayesian perspective leveraging priors inherited from phylogenetic literature to infer tree-based modular hierarchies across nodes, which explain heterogeneous multiscale patterns in the network. Identifiability results are derived along with posterior consistency theory. The inference potentials of our model are illustrated in simulations and two real-data applications from criminology and neuroscience, where our formulation learns core structures hidden to state-of-the-art alternatives.

stat.ME↗
arXiv2024

Saddlepoint Monte Carlo and its Application to Exact Ecological Inference

Assuming X is a random vector and A a non-invertible matrix, one sometimes need to perform inference while only having access to samples of Y = AX. The corresponding likelihood is typically intractable. One may still be able to perform exact Bayesian inference using a pseudo-marginal sampler, but this requires an unbiased estimator of the intractable likelihood. We propose saddlepoint Monte Carlo, a method for obtaining an unbiased estimate of the density of Y with very low variance, for any model belonging to an exponential family. Our method relies on importance sampling of the characteristic function, with insights brought by the standard saddlepoint approximation scheme with exponential tilting. We show that saddlepoint Monte Carlo makes it possible to perform exact inference on particularly challenging problems and datasets. We focus on the ecological inference problem, where one observes only aggregates at a fine level. We present in particular a study of the carryover of votes between the two rounds of various French elections, using the finest available data (number of votes for each candidate in about 60,000 polling stations over most of the French territory). We show that existing, popular approximate methods for ecological inference can lead to substantial bias, which saddlepoint Monte Carlo is immune from. We also present original results for the 2024 legislative elections on political centre-to-left and left-to-centre conversion rates when the far-right is present in the second round. Finally, we discuss other exciting applications for saddlepoint Monte Carlo, such as dealing with aggregate data in privacy or inverse problems.

stat.CO↗
arXiv2024

Insufficient Gibbs Sampling

In some applied scenarios, the availability of complete data is restricted, often due to privacy concerns; only aggregated, robust and inefficient statistics derived from the data are made accessible. These robust statistics are not sufficient, but they demonstrate reduced sensitivity to outliers and offer enhanced data protection due to their higher breakdown point. We consider a parametric framework and propose a method to sample from the posterior distribution of parameters conditioned on various robust and inefficient statistics: specifically, the pairs (median, MAD) or (median, IQR), or a collection of quantiles. Our approach leverages a Gibbs sampler and simulates latent augmented data, which facilitates simulation from the posterior distribution of parameters belonging to specific families of distributions. A by-product of these samples from the joint posterior distribution of parameters and data given the observed statistics is that we can estimate Bayes factors based on observed statistics via bridge sampling. We validate and outline the limitations of the proposed methods through toy examples and an application to real-world income data.

stat.ME↗
arXiv2023

Some are observed, all leave traces: whole-population modeling of French elite civil servants' career paths

Elite civil servants may come and go between the public and private sectors throughout their career, a process of particular interest for the public and social scientists. However, data to document such processes are rarely completely available: we need inference tools that can account for many missing values. We consider public-private paths of elite French civil servants and introduce binary Markov switching models with Bayesian data augmentation. Our procedure relies on two complementary data sources: (1) detailed observations of some individual trajectories obtained from LinkedIn; (2) less informative ``traces'' left by all individuals in the administrative record, which we model for missing data imputation. This model class maintains the properties of hidden Markov models and enables a tailored sampler to target the posterior, yet allows for varying parameters across individuals and time. By integrating the two sources, we can consider the whole population rather than just a sample, and avoid the biases that would stem from using only a single source. We demonstrate this allows to properly test substantive hypotheses on career paths across a variety of public organizations. We notably show that the probability for ENA graduates to exit the public sector has not increased since 1990, but that the probability they return has increased. We identify three clusters of organizations, with distinct patterns of public-private behaviors.

stat.AP↗
arXiv2023

Asymptotics of approximate Bayesian computation when summary statistics converge at heterogeneous rates

We consider the asymptotic properties of Approximate Bayesian Computation (ABC) for the realistic case of summary statistics with heterogeneous rates of convergence. We allow some statistics to converge faster than the ABC tolerance, other statistics to converge slower, and cover the case where some statistics do not converge at all. We give conditions for the ABC posterior to converge, and provide an explicit representation of the shape of the ABC posterior distribution in our general setting; in particular, we show how the shape of the posterior depends on the number of slow statistics. We then quantify the gain brought by the local linear post-processing step.

stat.CO↗
arXiv2023

TraitLab: a Matlab package for fitting and simulating binary tree-like data

TraitLab is a software package for simulating, fitting and analysing tree-like binary data under a stochastic Dollo model of evolution. The model also allows for rate heterogeneity through catastrophes, evolutionary events where many traits are simultaneously lost while new ones arise, and borrowing, whereby traits transfer laterally between species as well as through ancestral relationships. The core of the package is a Markov chain Monte Carlo (MCMC) sampling algorithm that enables the user to sample from the Bayesian joint posterior distribution for tree topologies, clade and root ages, and the trait loss, catastrophe and borrowing rates for a given data set. Data can be simulated according to the fitted Dollo model or according to a number of generalized models that allow for heterogeneity in the trait loss rate, biases in the data collection process and borrowing of traits between lineages. Coupled pairs of Markov chains can be used to diagnose MCMC mixing and convergence and to debias MCMC estimators. The raw data, MCMC run output, and model fit can be inspected using a number of useful graphical and analytical tools provided within the package or imported into other popular analysis programs. TraitLab is freely available and runs within the Matlab computing environment with its Statistics and Machine Learning toolbox, no other additional toolboxes are required.

stat.CO↗
arXiv2022

Lagged couplings diagnose Markov chain Monte Carlo phylogenetic inference

Phylogenetic inference is an intractable statistical problem on a complex space. Markov chain Monte Carlo methods are the primary tool for Bayesian phylogenetic inference but it is challenging to construct efficient schemes to explore the associated posterior distribution or assess their performance. Existing approaches are unable to diagnose mixing or convergence of Markov schemes jointly across all components of a phylogenetic model. Lagged couplings of Markov chain Monte Carlo algorithms have recently been developed on simpler spaces to diagnose convergence and construct unbiased estimators. We describe a contractive coupling of Markov chains targeting a posterior distribution over a space of phylogenetic trees with branch lengths, scalar parameters and latent variables. We use these couplings to assess mixing and convergence of Markov chains jointly across all components of the phylogenetic model on trees with up to 200 leaves. Samples from our coupled chains may also be used to construct unbiased estimators.

stat.ME↗
arXiv2022

A Phylogenetic Model of the Evolution of Discrete Matrices for the Joint Inference of Lexical and Phonological Language Histories

We propose a model of the evolution of a matrix along a phylogenetic tree, in which transformations affect either entire rows or columns of the matrix. This represents the change of both lexical and phonological aspects of linguistic data, by allowing for new words to appear and for systematic phonological changes to affect the entire vocabulary. We implement a Sequential Monte Carlo method to sample from the posterior distribution, and infer jointly the phylogeny, model parameters, and latent variables representing cognate births and phonological transformations. We successfully apply this method to synthetic and real data of moderate size.

q-bio.PE↗
arXiv2021

Evolution and trade-off dynamics of functional load

Function Load (FL) quantifies the contributions by phonological contrasts to distinctions made across the lexicon. Previous research has linked particularly low values of FL to sound change. Here we broaden the scope of enquiry into FL, to its evolution at all values. We apply phylogenetic methods to examine the diachronic evolution of FL across 90 languages of the Pama-Nyungan (PN) family of Australia. We find a high degree of phylogenetic signal in FL. Though phylogenetic signal has been reported for phonological structures, such as phonotactics, its detection in measures of phonological function is novel. We also find a significant, negative correlation between the FL of vowel length and of the following consonant, that is, a deep-time historical trade-off dynamic, which we relate to known allophony in modern PN languages and compensatory sound changes in their past. The finding reveals a historical dynamic, similar to transphonologization, which we characterize as a flow of contrastiveness between subsystems of the phonology. Recurring across a language family which spans a whole continent and many millennia of time depth, our finding provides one of the most compelling examples yet of Sapir's 'drift' hypothesis, of non-accidentally parallel development in historically related languages.

cs.CL↗
arXiv2018

Calibration procedures for approximate Bayesian credible sets

We develop and apply two calibration procedures for checking the coverage of approximate Bayesian credible sets including intervals estimated using Monte Carlo methods. The user has an ideal prior and likelihood, but generates a credible set for an approximate posterior which is not proportional to the product of ideal likelihood and prior. We estimate the realised posterior coverage achieved by the approximate credible set. This is the coverage of the unknown ``true'' parameter if the data are a realisation of the user's ideal observation model conditioned on the parameter, and the parameter is a draw from the user's ideal prior. In one approach we estimate the posterior coverage at the data by making a semi-parametric logistic regression of binary coverage outcomes on simulated data against summary statistics evaluated on simulated data. In another we use Importance Sampling from the approximate posterior, windowing simulated data to fall close to the observed data. We illustrate our methods on four examples.

stat.CO↗
arXiv2014

The Wang-Landau algorithm reaches the flat histogram criterion in finite time

The Wang-Landau algorithm aims at sampling from a probability distribution, while penalizing some regions of the state space and favoring others. It is widely used, but its convergence properties are still unknown. We show that for some variations of the algorithm, the Wang-Landau algorithm reaches the so-called flat histogram criterion in finite time, and that this criterion can be never reached for other variations. The arguments are shown in a simple context - compact spaces, density functions bounded from both sides - for the sake of clarity, and could be extended to more general contexts.

math.ST↗
arXiv2009

Missing data in a stochastic Dollo model for cognate data, and its application to the dating of Proto-Indo-European

Nicholls and Gray (2008) describe a phylogenetic model for trait data. They use their model to estimate branching times on Indo-European language trees from lexical data. Alekseyenko et al. (2008) extended the model and give applications in genetics. In this paper we extend the inference to handle data missing at random. When trait data are gathered, traits are thinned in a way that depends on both the trait and missing-data content. Nicholls and Gray (2008) treat missing records as absent traits. Hittite has 12% missing trait records. Its age is poorly predicted in their cross-validation. Our prediction is consistent with the historical record. Nicholls and Gray (2008) dropped seven languages with too much missing data. We fit all twenty four languages in the lexical data of Ringe (2002). In order to model spatial-temporal rate heterogeneity we add a catastrophe process to the model. When a language passes through a catastrophe, many traits change at the same time. We fit the full model in a Bayesian setting, via MCMC. We validate our fit using Bayes factors to test known age constraints. We reject three of thirty historically attested constraints. Our main result is a unimodel posterior distribution for the age of Proto-Indo-European centered at 8400 years BP with 95% HPD equal 7100-9800 years BP.

stat.AP↗