Synthetic Data¶
All analyses run on synthetic datasets generated by synthetic_data/generate_synthetic_data.py.
Synthetic data keeps the focus on the statistical concept rather than domain noise, makes every
result reproducible, and lets the notebooks construct exact edge cases.
scipymasterpro generate-data # writes to synthetic_data/exports/
scipymasterpro generate-data -o /tmp/out # or anywhere else
| File | Generator | Shape | Used by |
|---|---|---|---|
normal_skewed.csv |
generate_normal_skewed |
normal + skew-normal columns | 01, 02, S1 |
mixed_distributions.csv |
generate_mixed_distributions |
gamma, beta, exponential, lognormal columns | 03 |
multivariate_gaussian.csv |
generate_multivariate_gaussian |
3 correlated Gaussian columns | 06, 08 |
sample_for_optimization.csv |
generate_sample_for_optimization |
noisy quadratic | 07 |
curve_fitting_data.csv |
generate_noisy_curve_fitting_data |
noisy exponential decay | 09 |
poisson_data.csv |
generate_poisson_data |
Poisson counts | 04 |
grouped_continuous.csv |
generate_grouped_continuous |
three groups with different means | 02, 04 |
bootstrap_sample_data.csv |
generate_bootstrap_sample_data |
single normal sample | 05, 10 |
categorical_counts.csv |
generate_categorical_counts |
category frequency table | 04 |
Every generator takes a seed (default 42) so the CSVs are reproducible; the committed copies
in synthetic_data/exports/ are what the notebooks and app load by default.