Activity

Participants generate a synthetic Beta4 `pij`, inspect ICCs for specific instances, train Beta4, and check parameter recovery.

Browser Lab

This page includes executable Python cells directly in the site. You can construct the synthetic pij matrix and inspect ICCs directly in the browser runtime. When a notebook backend is connected, from birt import Beta4 is also available so the same page can train Beta4 and check parameter recovery.

Shared Helpers You Can Use

The activity notebook, Colab, and the in-page runtime can all use the same shared helpers. This section is designed to keep the focus on Beta4 itself: latent parameter sampling, bounded-response construction, ICC interpretation, and recovery.

from utils.handson import (
    beta4_expected_response,
    plot_beta4_family,
)

• beta4_expected_response lets you evaluate ICCs from specific Beta4 parameters.
• plot_beta4_family is useful when you want to compare several Beta4 parameter settings quickly.
• When the backend is connected, from birt import Beta4 is available for the training and recovery steps.

Inside the browser notebook editor, hovering one of these helper names now shows its docstring.

Notebook

• Activity notebook: 03_01_activities.ipynb
• Goal: understand Beta4 as a bounded-response model before the workshop moves to agreement-based evaluation

What Participants Should Do

• generate a synthetic bounded-response matrix pij from the parameter distributions shown in the section;
• generate ICCs for specific instances, such as the easiest and the hardest;
• train Beta4 on the synthetic pij and recover the latent parameters;
• compare the original and estimated parameters, especially through recovery plots such as theta_i versus theta_i_hat.