Working in Bayesian statistics, statistical modeling and scientific computing.
I come from a molecular biology background at Sofia University and have moved progressively deeper into statistical modeling and scientific computing. Most of my work sits at the point where a biological or physical question has to be turned into a model that can actually be fit, checked, and trusted — which usually means writing the model, writing the tooling around it, and writing the simulation study that tells me whether either one works.
I build Bayesian models for scientific and applied problems: hierarchical and multilevel structures for grouped experimental data, mechanistic and nonlinear models where the parameters carry physical meaning, causal models for questions that observational or trial data can't answer by regression alone, and spatial and time-series structure where the data demands it. Most of this is written in Stan and driven from R through reproducible targets pipelines.
Alongside the models, I develop the software that supports them — R packages for simulation, model validation, causal inference and visualization, and a preprocessor and language tooling that give Stan a package manager and namespaced imports it doesn't have natively.
|
diet-exercise-rct-t2d Bayesian analysis of a diet and exercise trial in type 2 diabetes, with the causal structure written as an explicit DAG and a simulated data-generating process for anthropometric outcomes. Bayesian Causal Inference DAGs Stan R targets
|
emf-maize-bayes Hierarchical Bayesian modeling of maize growth and physiology under 868 MHz electromagnetic field exposure, spanning Gompertz growth curves, cell-means models, JIP-test photosynthesis parameters and biochemistry endpoints. Hierarchical Bayes Gompertz Growth Stan R
|
|
University-Biomembranes- Bayesian analysis of electroinduced erythrocyte lysis, running from Stan models through a reproducible pipeline to a LaTeX write-up. Bayesian Biophysics Stan R targets
|
bayesian-viral-infectivity Bayesian estimation of the probability that a cell becomes infected, across a dilution series. Bayesian Dilution Series Virology R
|
|
Bayes-Langmuir-Vs-Classical-Langmuir Bayesian nonlinear regression for thermodynamic parameters from Langmuir monolayer compression isotherms, set against the classical fitting approach. Nonlinear Regression Thermodynamics Bayesian Stan
|
|
airbnb-listings- Hierarchical Bayesian price modeling for Airbnb listings, with simulation-based checks and a targets pipeline that can execute remotely on AWS.Hierarchical Bayes Stan R targets AWS
|
|
Laplace is a source-to-source preprocessor for Stan. It compiles .laplace files down to plain, readable .stan files, adding a package manager and namespaced pkg::func() imports to a language that has neither natively.It never becomes a runtime dependency: once build/model.stan exists, it can be handed to stanc or CmdStan with Laplace uninstalled.
|
|
laplace The compiler and package manager. Resolves and installs dependencies, builds .laplace sources into committable .stan files, and optionally validates them against stanc.Rust Stan Compilers Package Management
|
|
laplace-tools Editor tooling — a language server plus a VS Code extension giving block-role and function-origin coloring, autocomplete across imported libraries, and live diagnostics for unresolved imports and stale version pins. Language Server Rust TypeScript VS Code
|
transformation A Laplace package of data transformation functions for Stan: centering, scaling and standardization, Box-Cox and Yeo-Johnson, rank and quantile transforms. Laplace Package Stan Transformations
|
|
bsst Bayesian Simulation-based Severe Testing — an R package for validating Stan models through parameter recovery experiments, built to locate the sample sizes, effect sizes and designs where inference quietly breaks down. R Stan Parameter Recovery Model Validation
|
rsims An R package for simulating statistical datasets: hierarchical and crossed designs, random and correlated effects, splines, CFA/SEM helpers and state-space processes. R Simulation Hierarchical Data State-Space
|
|
tidycausality Meta machine-learning algorithms for causal inference in R — S-, T-, X- and R-learners built on the tidymodels framework. R Causal Inference Meta-Learners tidymodels
|
stanviz Reusable ggplot2-based plotting functions for fitted Stan models. R Stan ggplot2 Visualization
|
|
qcflowr An RNA-seq workflow from raw FASTQ to gene counts, automating quality control, trimming, alignment and quantification with FastQC, fastp, HISAT2 and featureCounts. Shell RNA-seq Bioinformatics Pipelines
|
Bayesian inference · hierarchical and multilevel models · causal inference · nonlinear and mechanistic models · state-space and time series · spatial modeling · simulation-based validation and parameter recovery · probabilistic programming · supervised learning and model explainability
R ; Stan ; Julia ; Python : Shell / Bash ; SQL ; LaTeX
FastQC · fastp · HISAT2 · featureCounts
Contact — metodilatinov@abv.bg