Continuous and mixed joints
Gaussian dependence, continuous margins, Clayton mixed measures, seeded sampling, diagnostics, and family selection.
Integration example →Fitting example →
Executable example center
Start from code that the build verifies
The catalog separates JDistlib scripts, ordinary .stan compatibility fixtures, and general Java APIs. Every example is exercised by ./gradlew check.
Stan source compatibility · expanded for 0.8.3
Forty-one ordinary .stan fixtures
These fixtures cover literals, multidimensional arrays and assignments, arrays of vectors/matrices, complex and tuple values, forward-declared, external, recursive, and higher-order functions, sparse/shaped matrix algorithms, structured constraints, broadcasting, probability kernels, quadrature, roots, stiff ODEs, and DAEs. They compile through ModelScript.compileStan and run on JDistlib rather than Stan's C++ runtime.
Stan-inspired modeling-language scripts · expanded in 0.8.x
Fifty-one Stan-inspired model scripts
This section—and only content explicitly carrying the same label—uses JDistlib's Stan-inspired script syntax. The catalog covers conjugate posteriors, regression links, robust and overdispersed likelihoods, control flow and locals, stable log-space mixtures, positive and bounded parameters, ordered vectors, simplexes, transformed blocks, indexing, vectorization, file-backed data, Wiener reaction times, and posterior-predictive RNGs.
General Java examples · not Stan-specific
Choose the Java workflow you need
These examples demonstrate ordinary JDistlib Java APIs. A card is labeled separately when it also includes the Stan-inspired script frontend.
Fitting example →
Weighted population models
Evaluate mixture density/CDF/quantiles and generate reproducible draws from multiple regimes.
Open example →Change support and units
Affine and nonlinear transforms, truncation, censoring, explicit inverse maps, and Jacobians.
Integration example →Composition examples →
Batch to online testing
BH, weighted BH, q-values, grouped selection, LORD++, SAFFRON, censored families, and discrete DBH.
Quick workflow →Full catalog →
Build from a kernel
Analyze formulas, normalize continuous and discrete laws, inspect diagnostics, compose a contamination model, and calculate intervals.
Integration example →Builder examples →
Java, CSV, scripts, and compilation
Start with a fully commented CSV-to-conclusion analysis, or use the smaller examples to compare Java/script frontends and compilation modes.
Read the complete walkthrough →Open the worked Java file →
Compare data frontends →
Compare compilation modes →
Compare Java/script MCMC →
15 tested models →
Covariate and locus selection
Build a complete normalized log joint, move through ragged models with add/drop/swap proposals, inspect inclusion mixing, export draws, and resume an exact checkpoint.
Read the worked analysis →Open the complete Java file →
17,000-gene mixed-model selection
Download GSE93272, prepare array expression and clinical covariates, fit a subject random-intercept target, accelerate residual-informed proposals, and resume crash-safe segments.
Read the public-data tutorial →Open the complete Java file →
Finite and infinite intervals
Integrate smooth finite functions, positive-tail laws, and whole-line kernels with explicit result diagnostics.
Open example →Count, simplex, Wishart, and first-passage events
Use exact count-vector DP, error-reporting continuous rectangles, named Wishart trace/directional/determinant events, and the Wiener reaction-time density and RNG.
Open example →Read the rectangle contract →
Read the Wishart contract →
Tails, transforms, aggregation, EVT, and options
Use checked tail/payoff functionals, tempered-stable and multivariate laws, exact or numerical aggregation, advanced risk and path functionals, implied volatility, and atom-aware or smooth option-implied distributions.
Read tutorial →Open core feature examples →
Open advanced examples →
Worked options analysis →
Open worked Java file →
Sampling smoke benchmark
Track vectorized gradient evaluation and NUTS transition throughput without mistaking a smoke run for a JMH benchmark.
Open benchmark →Roots, quadrature, ODEs, and DAEs
Java and script callbacks, propagated sensitivities, adaptive RK45/BDF, implicit index-1 DAEs, and projected holonomic index-3 mechanics.
Read tutorial →Open Java example →
Open script fixtures 38–41 →
Reusable tape and atomic kernels
Lower compiled scripts onto thread-local reusable tapes and compare automatic lowering with a hand-authored primitive target.
Read guide →Open benchmark →
Packaged-JAR compatibility
Compile and run a client against the release JAR on Java 8, independently of the build JVM.
Open smoke test →Run locally
Compile and validate the whole catalog
./gradlew check
./gradlew validateModelScripts
# Windows
gradlew.bat check
gradlew.bat validateModelScriptscheck compiles every Java example against the packaged library and invokes the model-script validator. The validator compiles all fifty-one JDistlib scripts and forty-one ordinary .stan fixtures with representative data or external bindings, checks their gradients, verifies finite initial densities, and executes generated quantities.
Learning paths
Examples are starting points, not substitutes for diagnostics
Use the learning center for conceptual tutorials, the inference guide for computational review, and the JavaDoc for complete signatures.