LM-hypotest 0.1.0

A Python library for hypothesis testing with automated assumptions

Hypotest

A Python library for deterministic hypothesis testing with automatic assumption checking and optional LLM-based interpretation.

Hypotest provides a clean statistical engine designed for data scientists, researchers, and engineers who need reliable and reproducible statistical testing workflows.

Overview

Hypotest simplifies hypothesis testing by providing:

A deterministic statistical engine

Automatic assumption validation (normality, variance homogeneity)

Structured result objects with statistical metadata

Optional LLM-based interpretation layer

A safe Dataset abstraction for robust data handling

All statistical computations are deterministic and independent of LLM usage.

Installation

Once published:

pip install hypotest

Development install:

git clone https://github.com/chikku1234568/Unified-EDA-HypoTest-LM-Library cd hypotest pip install -e .

Optional LLM support:

pip install hypotest[llm]

Quick Start Example: Independent t-test import pandas as pd import numpy as np

import hypotest from hypotest.core.dataset import Dataset from hypotest.tests.parametric.ttest import TTest

Create example dataset

df = pd.DataFrame({ "group": ["A"] * 100 + ["B"] * 100, "value": np.concatenate([ np.random.normal(0, 1, 100), np.random.normal(1, 1, 100), ]) })

Wrap DataFrame in Dataset abstraction

dataset = Dataset(df)

Run t-test

test = TTest()

result = test.execute( dataset=dataset, target="value", features=["group"], )

print(result)

Output:

TestResult(test='Independent t-test', feature='group', statistic=4.231, p=0.00003, significant)

Automatic Assumption Checking

Hypotest automatically checks statistical assumptions before or during test execution.

for assumption in result.assumptions: print(assumption.assumption_name, assumption.passed)

Example output:

normality True homoscedasticity False

Each assumption provides:

statistical result

interpretation

recommendation

Optional LLM Interpretation

Hypotest can generate natural-language explanations using any OpenAI-compatible provider.

Example using DeepSeek:

hypotest.configure( llm_api_key="your-api-key", llm_base_url="https://api.deepseek.com/v1", llm_model="deepseek-chat", enable_llm_interpretation=True, )

print(result.explain())

Example output:

The independent t-test indicates a statistically significant difference between the two groups...

LLM interpretation is optional and does not affect statistical computation.

Configuration

Configure hypotest globally:

hypotest.configure( llm_api_key="your-key", llm_base_url="https://api.deepseek.com/v1", llm_model="deepseek-chat", enable_llm_interpretation=True, )

View configuration:

print(hypotest.info())

Dataset Abstraction

Hypotest uses a Dataset wrapper to provide safe data handling:

from hypotest.core.dataset import Dataset

dataset = Dataset(df)

This enables:

safe missing value handling

validation before test execution

future extensibility

Supported Tests (Current MVP)

Independent t-test

Planned:

Welch's t-test

Mann-Whitney U test

ANOVA

Chi-square test

Correlation tests

Features

Core features implemented:

Deterministic statistical engine

Automatic assumption checking

Structured TestResult objects

Dataset abstraction layer

Plug-in test registry system

Optional LLM interpretation

Planned features:

Automatic test recommendation

Effect size library

Automated reporting

Additional statistical tests

Example: Full Workflow import pandas as pd import numpy as np import hypotest

from hypotest.core.dataset import Dataset from hypotest.tests.parametric.ttest import TTest

hypotest.configure(enable_llm_interpretation=False)

df = pd.DataFrame({ "group": ["A"] * 50 + ["B"] * 50, "value": np.random.randn(100), })

dataset = Dataset(df)

test = TTest()

result = test.execute(dataset, "value", ["group"])

print(result)

for a in result.assumptions: print(a.assumption_name, a.passed)

print(result.explain()) # None if LLM disabled

Project Structure hypotest/ ├── core/ │ ├── dataset.py │ ├── result.py │ ├── tests/ │ ├── parametric/ │ ├── ttest.py │ ├── assumptions/ │ ├── normality.py │ ├── variance.py │ ├── llm/ │ ├── client.py │ ├── interpreter.py │ ├── config/ │ ├── manager.py │ ├── info.py

Requirements

Python ≥ 3.10

pandas ≥ 1.5

numpy ≥ 1.21

scipy ≥ 1.9

Optional:

openai-compatible client (for LLM interpretation)

Philosophy

Hypotest separates:

Deterministic statistical computation

Probabilistic natural-language interpretation

This ensures statistical correctness while enabling explainability.

License

MIT License