# Data — T2-simulation

This project uses **no observed data**. The only input is `design.csv`, written by the task designer: the nine cells of the Monte Carlo design (one row per cell).

| column | meaning |
|---|---|
| `cell_id` | C1 … C9 |
| `G` | number of clusters (5, 10, 30) |
| `icc` | intra-cluster correlation of the data-generating process (0.0, 0.2, 0.5) |
| `cluster_sizes` | the G cluster sizes n_1, …, n_G, separated by `|` (the pattern 8, 16, 32, 64 repeated) |
| `N` | total number of observations in a sample |
| `R` | number of Monte Carlo replications per cell (2000) |
| `B` | number of wild-bootstrap draws per replication (399) |
| `seed` | the seed to use for the cell (numpy `default_rng(seed)` or equivalent) |
| `mu` | population mean of y (0.0) |
| `total_variance` | Var(y_ig) = Var(u_g) + Var(e_ig) = 1.0 |
| `nominal_level` | nominal confidence level (0.95) |

Data-generating process for a cell: y_ig = mu + u_g + e_ig, g = 1..G, i = 1..n_g, with u_g ~ N(0, icc) independent across clusters and e_ig ~ N(0, 1 − icc) independent across observations and of u. All simulated data are synthetic; nothing here is observed data.