No packages match

proxymix - Kullback-Leibler Optimal Gaussian Mixture Proxies for Target Densities

Fits multivariate Gaussian-mixture proxies that are Kullback-Leibler optimal to user-supplied target densities on real Euclidean space. Three fitting regimes are unified under one verb: (i) closed-form moment matching for a single component, (ii) classical expectation-maximisation when independent samples are available, and (iii) importance-sampled expectation-maximisation that minimises the Kullback-Leibler divergence when the target can be evaluated point-wise but not (cheaply) sampled. Closed-form Gaussian-mixture operators (density, sampling, marginalisation, conditioning, divergence) round out the toolkit. The conditioning operator drives multiple imputation of data missing at random, covering the multimodal and heteroscedastic cases a single-Gaussian model cannot represent. Implements the regime hierarchy of van der Hoek and Elliott (2024) <doi:10.1080/07362994.2024.2372605>.

Last updated

5.86 score 11 downloads

PESTO - Parameter Estimation Optimised, with APSIM Coupling

High-performance parameter estimation, uncertainty quantification, and inverse modelling toolkit built on modernised PEST++ algorithms. Provides native R interfaces to iterative ensemble smoothers (IES), Gauss-Levenberg-Marquardt (GLM) solvers, global sensitivity analysis, and multi-objective optimisation under uncertainty. Includes surrogate-accelerated IES via Gaussian Process and Random Fourier Features, adaptive SVD backends (randomised SVD, LAPACK, Eigen), and convergence-aware adaptive ensemble sizing. Designed for large-scale environmental, hydrological, and agricultural models with support for highly-parameterised problems (>100,000 parameters). The toolkit is model-independent -- any R function can serve as the forward model -- with first-class in-process coupling to the APSIM agricultural-systems model via the 'apsimx' package as its flagship simulator partner.

Last updated

openblascpp

5.24 score 1 dependents 11 scripts

kernR - Kernel-Based Causal Distributional Testing

Kernel-based hypothesis tests for causal inference and distributional treatment effects. Implements backdoor-adjusted HSIC (bd-HSIC) for testing causal association, and doubly robust kernel statistics (DR-DATE, DR-DETT) for testing distributional treatment effects beyond mean shifts. Supports binary, continuous, and mixed treatments, hierarchical/nested data structures, and scales to large datasets via Nystrom approximation and random Fourier features. This package depends on 'PESTO' (GPL (>= 3)); kernR's own sources are released under the MIT licence, but the installed combination with 'PESTO' is a combined work subject to the terms of the GPL (>= 3).

Last updated

openblascpp

4.85 score

kalmix - General State-Space Modelling, Filtering and Change-Point Detection

A compact, dependency-light toolkit for linear-Gaussian and discrete-state state-space models. Provides a vector-state Kalman filter and Rauch-Tung-Striebel smoother in closed form, a native Gaussian-mixture (regime-switching) filter, hidden Markov model filtering and Viterbi decoding, a likelihood-ratio change-point detector, and a minimal single-case interrupted-time-series causal contrast built on the smoother. Mean-reverting spread estimation for pairs trading is included as one documented application of the general machinery, building on the scalar Ornstein-Uhlenbeck spread model of Elliott, van der Hoek and Malcolm (2005) <doi:10.1080/14697680500149370>. Methods follow Kalman (1960) <doi:10.1115/1.3662552>, Rauch, Tung and Striebel (1965) <doi:10.2514/3.3166> and Durbin and Koopman (2012) <doi:10.1093/acprof:oso/9780199641178.001.0001>.

Last updated

4.40 score

masque - Structurally Faithful Development Surrogates for Tabular Data

Turns a confidential tabular dataset - a single table, a folder of files, or a multi-sheet workbook - into a structurally faithful synthetic clone suitable for pipeline development. Experimental-design allocation and the NA pattern are preserved exactly; categorical design, treatment, and covariate vocabularies can be reversibly aliased; outcome and numeric-covariate values are re-simulated via a Gaussian copula that preserves the global covariance structure. Columns shared across tables are aliased consistently so the synthetic tables still join. A private 'recipe' object round-trips a pipeline written against the synthetic clone onto the original data. Not a differential-privacy or anonymisation tool: outputs are development surrogates, not public-release-safe artefacts.

Last updated

4.38 score

optimix - A Unified Contract and Auto-Selector for Efficient Optimisers

Provides a single typed problem contract over a curated portfolio of efficient optimisers, with an automatic per-instance selector that picks the best optimiser for a given problem and returns it in an 'stats::optim()'-compatible result object. The base solver runs with no suggested packages installed; each additional engine, and an optional mixture-valued mapper from 'proxymix', is loaded lazily only when used.

Last updated

4.26 score 1 scripts

decideR - Bayesian Decision-Theoretic Actions under Loss and Hard Constraints

Maps a posterior distribution or an 'orchestra_manifest' inference result to an optimal action under an explicit loss function and a hard feasibility constraint, returning the action together with a grounding label and an abstention when evidence is insufficient. The grounding label is combined worst-case across inputs, so a decision resting on an unverified external fact is downgraded to abstention rather than asserted with false confidence. A manifest-native tail closes an analytics pipeline by deciding directly from an upstream member's inference result, and a grade-band loss prices quality-threshold ('protein-band') payoffs as a first-class step value schedule. Reference decisions cover bounded input-rate selection and constraint-capped position sizing.

Last updated

4.02 score 1 dependents 9 scripts

cdzoo - A Multi-Backend Benchmark Harness and Recommender for Causal Discovery

Provides a single typed interface over a curated portfolio of causal-structure-discovery algorithms drawn from several R and Python back-ends, together with a structural-causal-model scenario grid, fair graph-recovery metrics, and an evidence-routed recommender that inspects a data set and suggests a discovery method suited to its characteristics. Every back-end is optional and loaded lazily only when used, so the harness, the scenario grid, the metrics, and the recommender's routing logic all run with no suggested package installed.

Last updated

3.85 score

orchestraManifest - The Crop Analytics Orchestra Manifest and Refusal Contract

A versioned, integrity-hashed and provenance-tracked 'S7' contract for the results that crop-analytics components exchange. A manifest records what was inferred, which component emitted it, the method and seed behind it, the load-bearing payload, and the lineage of any upstream manifest it was derived from; a 'SHA-256' hash over version-stable serialised bytes lets any consumer verify that the payload has not been modified since emission. Consumers declare the manifest versions they accept and refuse anything else loudly rather than coercing silently across a schema boundary. Companion predicates recognise a structured refusal or abstention from any component regardless of its object system, so a gate that should stop on a declined result cannot sail through it, and grounding tokens propagate worst-case so that one unverified input taints a downstream verdict. Adapters lift a 'PESTO' ensemble manifest and a 'proxymix' Gaussian mixture fit into the contract.

Last updated

3.78 score 1 dependents

conductoR - Perform a Contract-Typed Analytical Plan Across Analytics Components

Declares an analytical plan as a directed acyclic graph of nodes, each bound to a tool that returns an 'orchestraManifest' result, and performs it in topological order. Every edge is typed by the inferential target the upstream node declares, and a mismatched handoff stops the run before the downstream node executes rather than coercing across a contract boundary. Two epistemic operators are built in: a power-aware abstention gate that refuses to pass a null verdict the design could not have resolved, and a triangulation operator that reconciles any number of independent lenses on one question. A plan serialises to 'YAML' and back, compiles to a 'targets' pipeline, and every run writes a provenance record with the integrity chain of the manifests it produced.

Last updated

3.72 score 26 scripts

terroir - Geo-Point Environmental and Agricultural-Economics Data Collector

A uniform, point-keyed interface for retrieving climate, weather, vegetation-index, soil, elevation and agricultural-economics data for a geographic coordinate from many free data sources. Each source is wrapped as an optional adapter behind one request and response contract with mandatory provenance; the package builds and checks with no source package installed. The data-collector member of the ORCHESTRA.

Last updated

3.65 score

apsimR - APSIM Next Generation as a Contract-Emitting Analysis Member

A typed, server-capable R interface to APSIM Next Generation, the process-based agricultural systems simulator. Drives the simulator for prediction over scenario designs, mechanism-parameter estimation, parameter calibration against observations (a bounded optimiser, or an iterative ensemble smoother when 'PESTO' is present), global sensitivity analysis (Morris and Sobol), Gaussian-process emulation with leave-one-out diagnostics, validation against observations (goodness-of-fit metrics and a typed verdict), and the generation of known-causal-structure datasets for testing causal-inference methods, and returns results as a versioned, provenance-complete 'apsim_manifest' object that is emitted natively into the shared orchestra manifest contract, so simulator output composes directly with downstream calibration, uncertainty-quantification and causal-inference tools; an ensemble inversion additionally bridges to a 'PESTO' ensemble manifest. Execution degrades gracefully: every entry point abstains with a typed reason when the simulator or its runtime is absent.

Last updated

3.62 score 12 scripts

gpfield - Change-of-Support Gaussian-Process Regression for Field and Multi-Environment-Trial Data

A self-contained, contract-emitting member of the agricultural analytics orchestra for spatial and spatio-temporal Gaussian-process regression over field and multi-environment-trial layouts. Fits an exact Gaussian process to spatial (and optionally temporal) coordinates with a choice of squared-exponential or 'Matern' covariance, estimating the hyperparameters by marginal-likelihood optimisation. Its defining feature is change of support: a prediction can be requested at a support different from the one observed -- a point process predicted as block (paddock-scale) averages, or block averages disaggregated toward points -- with the predictive variance propagated through the support map. Predictions abstain with a typed, machine-readable reason when the estimated spatial range or the data support cannot bear the requested resolution, so the member refuses cleanly rather than returning a confident wrong surface. Results are emitted as a versioned, provenance-complete object compatible with the orchestra ensemble-manifest contract, so a smoothed field surface composes directly with downstream calibration, uncertainty-quantification, causal-inference and decision tools. The package is released under the MIT licence and takes no heavy external Gaussian-process dependency: the exact solver is base R.

Last updated

3.60 score

koine - Synthesised Fourth Opinion for Bayesian Triangulation

Synthesises one coherent, independent predictive opinion for a Bayesian model from a curated battery of known and underutilised inference methods, or abstains when no diversity gain exists. Built to serve as a fourth, synthesised witness alongside an independent panel of inference engines: it stacks over model-structure and method diversity in predictive space (leave-one-out stacking, PSIS-gated) on an axis orthogonal to the engine axis, never averaging the engines. Predictive opinions are exchanged through an ArviZ-style 'DataTree' 'NetCDF' interchange so no live 'Python' is required at read time. Abstention is a first-class outcome.

Last updated

3.60 score

grainPlan - Grain-Specific Decision Orchestration on Top of Decision Theory

Turns an orchestra inference result into an actionable grain-production plan. It is a thin grain-semantics layer on top of the 'decideR' decision engine: it reuses that engine, its grade-band loss library, its manifest-native pipeline tail, and its grounding firewall, and adds the grain economics those decisions kept re-deriving. Named grain decisions price a nitrogen rate by expected profit, rank a variety from a multi-environment trial under downside-aware merit, and target a quality grade against a delivery standard. A reusable, documented grain value library supplies wheat protein bands, barley malting premiums, and nitrogen-rate economics built from published price structures. A season plan composes several grain decisions into one provenance-carrying object whose grounding is the worst case across its decisions, so a plan resting on any unverified input abstains rather than asserting a confident wrong recommendation.

Last updated

3.54 score

survkit - A Unified, Extensible Toolkit for Time-to-Event Analysis

One clean interface over the established survival stack (parametric accelerated-failure-time models, Cox proportional hazards, competing risks, recurrent events, frailty, flexible splines and a non-parametric Kaplan-Meier baseline), driven by an extensible method registry so new estimators slot in without an API change. Fits return a single typed object with consistent prediction, summary and tidying methods. On top of the fits sit the value layer: cross-method triangulation that reads disagreement as a misspecification signal, restricted mean survival time and cumulative incidence, inverse-probability-of-censoring-weighted predictive scoring, a small-sample power gate that abstains when events are too few to identify the model, a registry conformance contract, and an optional emitter to an ensemble-manifest contract for composition with a wider analytics stack.

Last updated

3.48 score 2 scripts

picMort - Reproducible Pediatric ICU Mortality Benchmark on PIC V1.1.0

Tools and pipelines for a calibration-first pediatric ICU mortality benchmark on the open-access PIC v1.1.0 database. Authoritative cohort specification, T+24h prediction-window feature extraction, PIM3 reconstruction, calibration and decision-curve evaluation. The package is the methodological substrate for the companion manuscript, illustrated end to end in a worked vignette.

Last updated

3.48 score

bacipair - BACI / Paired-DiD / Synthetic-Control Estimators for Carbon-MRV Audits

A reproducibility-by-construction toolkit for Before-After-Control-Impact (BACI) audits of carbon-MRV registers. Provides a staggered-treatment simulator with a known oracle ATT, a paired difference-in-differences estimator with proponent-cluster-robust standard errors, a time-varying synthetic-control comparator with simplex-constrained donor weights (Wang and Carreira-Perpinan, 2013), a proponent-level cluster bootstrap, a Goodman-Bacon (2021) staggered-DiD decomposition wrapper, and 'targets'-skeleton helpers for end-to-end pipeline reproducibility. Designed against the Macintosh et al. (2024) Australian Human-Induced Regeneration setting but applicable to any paired BACI register.

Last updated

3.18 score