Single-replication simulation references — MS3789

Each of the 16 batch directories contains the exported partial-summary CSV(s)
for task 1, using 20,000 MCMC iterations and discarding the first 4,000.
DGP_A/A1–A3 use N=50 and T=30,60,120 respectively; A4 runs N=[100,200]
and T=[30,60,120] together. benchmark/L1–L9 follow the nine singleton
combinations in README.pdf. smallT uses N=50, T=[10,15,20].
Robustness uses N=50, T=30; industry_network uses N=1616, T=32.

Follow README.pdf for running task 1 into a new directory and collecting
partial summaries. Compare parameter labels, posterior means, posterior
standard deviations and available interval summaries. Runtime columns and
small numerical differences can vary across environments. These are not
500-replication averages. Monte Carlo SD ratios and coverage performance
cannot be meaningfully assessed from one replication.

Only aggregate CSV outputs are included. Raw MAT files, input networks and
local run logs are not included here. The industry-network input remains
restricted; its reference CSV does not provide access to that network.
