What is the Synthetic Data Module?

The Synthetic Data Module generates privacy-safe "digital twins" of real datasets. You provide your data (or just its statistical summary), and the module produces a brand-new synthetic dataset that preserves the original's statistical structure -- distributions, correlations, and effect sizes -- without containing a single one of your original data points.

The result is a dataset you can share, publish, and analyse freely -- for collaboration, teaching, machine-learning training, or methods development -- without exposing confidential or regulated information.


What "synthetic data" means here

Synthetic data mirrors the statistical patterns of a real dataset without reusing any actual records. Because no original observation survives into the output, the synthetic dataset can be distributed without the privacy, consent, and regulatory constraints that apply to the source data. The statistical relationships -- how variables are distributed, how they correlate, how large the effects between them are -- are reconstructed so that analyses reach the same conclusions they would on the original.


The four-step workflow

Step What happens
1. Upload Your Data Provide an Excel file (or enter summary statistics manually)
2. Automatic Analysis The module detects each variable's type, distribution, and relationships
3. Configure Settings Set target sample size, review correlations and effect sizes, fine-tune parameters
4. Generate & Validate The engine produces the synthetic dataset and a full validation report in seconds

What the module preserves

Capability What it does
Privacy Protection Synthetic dataset can be shared without exposing any original data point
Statistical Integrity Correlation structures preserved using Cholesky decomposition
Effect Size Matching Cohen's d, Odds Ratios, and Hazard Ratios are maintained
Distribution Fitting Variable distributions (Normal, Uniform, Binomial, etc.) detected and replicated
Visual Validation Love plots, correlation matrices, and SMD comparisons produced every run
Fast Processing Optimised algorithms generate well over 100,000 rows in seconds

Who the module is for

Audience Typical uses
Clinical Research Power-analysis calculations, method development, IRB-compliant datasets, multi-site studies
Machine Learning Training-data augmentation, class-imbalance handling, model validation, feature engineering
Education Statistics coursework, workshop demonstrations, student exercises, publication examples

How accurate is the generated data?

The module preserves the large majority of a dataset's statistical relationships, and every generated dataset is accompanied by a validation report showing standardised-mean-difference (SMD) analysis and quality scores, so you can verify the fidelity of each run.


Supported data

The module supports six variable types -- continuous, binary, categorical, count, time-to-event, and proportion -- and detects each type automatically. Datasets can range from a few hundred rows to well over 100,000 rows in a single run.


Data sources

Source When to use
Raw data upload You have the original dataset as an Excel file
Manual input You want to define variables and parameters by hand
Descriptive statistics You only have a published statistical summary (means, SDs, correlations)

The third option is useful when you want to generate a dataset matching summary statistics reported in a paper without accessing the underlying records.


Privacy and data handling

The module is built privacy-first: uploads are encrypted, and the original data is used only to analyse its statistical structure -- it is not retained after processing. Because the synthetic output contains no original records, it sidesteps the confidentiality and regulatory constraints that govern the source data.


Access states

State What you see
Not signed in Redirected to the sign-in page with a return link
Signed in, no module access Redirected to the dashboard (module requires active subscription)
Signed in, with active access Full generation workflow