Sign In
Statistical Workstation

Rigorous statistics, conversational pace

A dedicated statistical reasoning engine for end-to-end inference. Describe your study; the engine selects the right methods, runs production-grade implementations, and attaches full provenance to every figure, table, and number — so your results reproduce byte-for-byte and reviewers can audit them end-to-end.

Sign In
data_prep.log
# raw_REDCap_export.csv
id,age,blood_press,date_str
001, 45 , "120/80", 2026-05-12
002, NULL, "140/90", 2026-05-13
⬇️ Ingest & Sanitize
# verified_dataset.parquet (Type-safe schemas)
[✓] Strip whitespace from numeric cells
[✓] Coerce missing values (NULL -> NaN)
[✓] Sniffed ';' delimiters & parsed date ISO-strings

Zero-effort data prep

Upload messy CSVs or Excel sheets — REDCap exports, lab dumps, survey extracts, whatever shape they arrive in. An ephemeral data prep stage sniffs delimiters, coerces standard formats, flags outliers, and records a deterministic cleaning recipe so the next file of the same shape replays in one click.

analysis_session.log
"Plot the survival curve of target by drug treatment, add the risk table."
"Fitting Kaplan-Meier model via survival::survfit. Plot generated."
🔬 Plot saved as artifact_id=abc123d
DPI: 300 Format: PNG / SVG / PDF

End-to-end inference, in plain language

Describe the study in plain language — t-tests, Cox proportional-hazards models, longitudinal mixed-effects fits, propensity matching, or anything in the statistical canon. A dedicated reasoning engine selects the right method, runs production-grade implementations under the hood, and surfaces results with audit-ready provenance attached.

provenance_audit.json
raw_data.csv
analysis_turn_5
manuscript_table.docx
[✓] Audit-Ready Provenance

"Computed via R 4.5.0 and survival (v3.7-0). deterministic seed=1735689712 was utilized."

Reproducible by construction

Every figure, table, and summary records its exact package versions, the underlying analysis code, locale state, and random seed. Export self-contained reproducibility bundles for re-runs, archives, or hand-off to a collaborator — no manual environment curation, no missing dependencies.

Built for research that has to hold up

Feature gauss.bio RStudio / R-stats ChatGPT + Data
**Conversational UI** Yes (Stateful stats chat) No (Coding only) Yes (Permissive)
**Publication-grade Output** Yes (Flextable DOCX / PDF / SVG) Yes (Via Quarto) No (Plain text / raw images)
**Verified Reproducibility** Yes (Captures package versions + seeds) Yes (Manual renv) No (Stochastic Python env)
**Data Privacy** Yes (HIPAA-fenced models) Yes (Local computation) No (Default open training)
**Audit Lineage** Yes (Graph of dependencies) No (Manual history log) No (Volatile session state)
🏥

Clinical & Epidemiological

Build Table 1 summaries with standardized mean differences, fit Cox proportional-hazards models with diagnostic checks, and export editable DOCX tables for publications, regulatory submissions, or internal reports.

🧠

Behavioral & Survey Research

Verify survey reliability, work with multi-construct Likert variables, handle item-level missingness, evaluate ANOVA multiplicity with Holm or Tukey corrections, and fit linear mixed models for nested or repeated designs.

🌿

Life Sciences & Ecology

Work with repeated-measures count matrices, inspect missing-data patterns with interactive missingness plots, apply multiple imputation, and produce high-resolution figures fit for publication or technical reports.