> ## Documentation Index
> Fetch the complete documentation index at: https://sourcetable.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Data science overview

> Run data science workflows in your spreadsheet with Python, ML, and statistics — no environment setup required.

Sourcetable gives you a complete data science environment inside your spreadsheet. Ask the AI to run analyses in plain English and it executes Python code with pandas, scikit-learn, StatsModels, and more — no installs, no notebooks, no configuration.

## How it works

<Steps>
  <Step title="Prepare your data">
    Import data via file upload, connector, or paste. Sourcetable handles CSV, XLSX, Parquet, JSON, and 15+ other formats.
  </Step>

  <Step title="Ask the AI">
    Describe what you want in the chat: "Run a correlation analysis on columns A through F" or "Build a classification model to predict churn."
  </Step>

  <Step title="Review results">
    The AI writes and executes Python code, then writes results back to your spreadsheet. You can view, modify, and re-run the generated code.
  </Step>
</Steps>

## Available libraries

| Library           | Capabilities                                                                        |
| ----------------- | ----------------------------------------------------------------------------------- |
| **pandas**        | Data manipulation, cleaning, aggregation, merging, time series                      |
| **NumPy**         | Numerical computing, linear algebra, random sampling                                |
| **SciPy**         | Statistical tests, optimization, signal processing, interpolation                   |
| **StatsModels**   | Regression, ANOVA, time series (ARIMA/SARIMAX), hypothesis tests                    |
| **scikit-learn**  | Classification, regression, clustering, dimensionality reduction, feature selection |
| **TabPFN**        | Zero-shot predictions and classifications without training                          |
| **plotly**        | Interactive charts and visualizations                                               |
| **matplotlib**    | Static plots, histograms, heatmaps                                                  |
| **seaborn**       | Statistical visualizations, distribution plots, pair plots                          |
| **bokeh**         | Interactive browser-based visualizations                                            |
| **transformers**  | NLP, text classification, sentiment analysis                                        |
| **BeautifulSoup** | Web scraping and HTML parsing                                                       |
| **SQLAlchemy**    | Database queries from Python                                                        |

## What you can do

<Columns cols={2}>
  <Card title="Exploratory data analysis" icon="magnifying-glass-chart" href="/data-science/eda">
    Profile datasets, find distributions, and surface patterns automatically.
  </Card>

  <Card title="Statistical analysis" icon="chart-scatter" href="/data-science/statistics">
    Hypothesis tests, regression, ANOVA, and correlation analysis.
  </Card>

  <Card title="Machine learning" icon="brain" href="/data-science/machine-learning">
    Classification, regression, clustering, and zero-shot predictions with TabPFN.
  </Card>

  <Card title="Forecasting" icon="chart-line-up" href="/data-science/forecasting">
    ARIMA, exponential smoothing, and ML-based time series forecasting.
  </Card>

  <Card title="Anomaly detection" icon="triangle-exclamation" href="/data-science/anomaly-detection">
    Statistical and ML methods to find outliers, fraud, and quality issues.
  </Card>

  <Card title="Feature engineering" icon="gears" href="/data-science/feature-engineering">
    Transform raw data into model-ready features with encoding, scaling, and creation.
  </Card>
</Columns>

## Who it's for

* **Analysts** — Run statistical tests and build models without leaving your spreadsheet
* **Data scientists** — Prototype faster with instant Python execution and visualization
* **Researchers** — Process experimental data with proper statistical methods
* **Business users** — Get data science results through natural language, no coding required