# Estimator Box

```{figure} ../../_static/images/tetrad-interface/box-by-box/estimator-box.png
:name: tetrad-estimator-box-screenshot
:alt: Estimator Box in the Tetrad interface.

Estimator Box in the Tetrad interface sidebar and main panel.
```

## Purpose

The **Estimator** box is where you **fit parametric models to data**. It connects:

- a **Parametric Model** (from the *Parametric Model* box),
- a **dataset** (from the *Data* box),
– and a choice of estimation method (e.g., ML, Dirichlet, EM, GLS, or robust variants, depending on the model type),

and produces **parameter estimates, standard errors (when available), and fit statistics**. The fitted result
can then be stored as an **instantiated model**.

## Typical workflow

1. **Choose a parametric model and dataset**
   - In the *Parametric Model* box, define the model structure and parameters.
   - In the *Data* box, verify that variable names and types match the model.
   - In the Estimator box, select:
     - The parametric model to be estimated.
     - The dataset on which to estimate it.

2. **Select an estimation method**
   - From the Estimator configuration panel, choose the relevant estimator (for example):
     - Maximum likelihood (ML).
     - Weighted least squares or robust estimators (if available).
   - Adjust any estimator-specific options (e.g., tolerance, maximum iterations, missing-data handling).

3. **Run the estimator**
   - Click **Run** to estimate the model.
   - Progress and any warnings or errors are typically reported in a log or message area.

4. **Inspect results**
   - Once estimation finishes, inspect:
     - Parameter estimates and (when supported) standard errors.
     - Fit indices (e.g., χ², RMSEA, CFI, BIC).
     - Convergence status and diagnostics.
   - If you are satisfied, save or register the result as an **instantiated model**.

5. **Reuse the fitted model**
   - Use the instantiated model in:
     - *Simulation* (to generate synthetic data from the fitted model),
     - *Compare* (to compare fits across different models),
     - or other tools that require fully specified, data-tied models.

## Key controls

- **Toolbar**
  - **New / Configure** – set up a new estimation task or modify an existing one.
  - **Run** – start estimation using the current settings.
  - **Stop** – interrupt a long-running estimation.
  - **Save / Instantiate** – create an instantiated model from the last successful fit (depending on version).
  - **Export** – save parameter estimates and fit statistics to a file, when supported.

- **Estimation setup panel**
  - Drop-downs or selectors for:
    - Parametric model.
    - Dataset.
    - Estimation method.
  - Additional options for:
    - Missing data handling.
    - Convergence criteria.
    - Robustness or scaling options (when available).

- **Results panel**
  - A table of parameter estimates and possibly:
    - Standard errors.
    - p-values or confidence intervals.
  - A summary of model fit:
    - χ², df, p-values.
    - RMSEA, CFI, TLI, BIC, etc., if provided by the estimator.
  - Warnings about convergence or identification problems.

## Common patterns & tips

- Always confirm that **variable names and ordering** in the parametric model match those in the dataset.
- If estimation fails or gives suspicious results:
  - Check for identification issues in the model.
  - Inspect the data for outliers, missingness patterns, or collinearity.
  - Try a different estimator or adjust convergence settings.
- When comparing models, keep separate estimation runs (and instantiated models) with descriptive names
  indicating the estimator used and key options.

## Estimator types and detail pages

The exact options available in the **Estimator** box depend on the type of **Parametric Model** connected to it.
Use the links below to see the detail pages for each estimator.

| Parametric model type | Estimator option            | Detail page                                      |
|-----------------------|-----------------------------|--------------------------------------------------|
| Bayes PM              | ML Bayes Estimator          | `Tetrad Interface → ML Bayes Estimator`          |
| Bayes PM              | Dirichlet Estimator         | `Tetrad Interface → Dirichlet Estimator`         |
| Bayes PM              | EM Bayes Estimator          | `Tetrad Interface → EM Bayes Estimator`          |
| SEM PM                | SEM Estimator               | `Tetrad Interface → SEM Estimator`               |
| Hybrid CG PM          | Hybrid CG Estimator         | `Tetrad Interface → Hybrid CG Estimator`         |
| Generalized SEM PM    | Generalized SEM Estimator   | `Tetrad Interface → Generalized SEM Estimator`   |

## Related pages

- `Tetrad Interface → Overview` – high-level tour of the GUI.
- Other boxes that commonly interact with **Estimator**:
  - *Parametric Model* (provides the model specification).
  - *Data* (provides the dataset to be fit).
  - *Instantiated Model* (stores the fitted model).
  - *Simulation* (can use fitted models as generative mechanisms).
  - *Compare* (compare fits or parameters across different estimation runs).