Replication data and code for the simulations in \"Quantile Ratio Regression for Panel Data\"

The scripts simulate both balanced (T = 2) and unbalanced (T in {4, 6}) panel data for n = 5000 units. The data generation process accounts for: Covariate Correlation: Two regimes (rho in {0, 0.5}). Conditional Distributions: Log-normal and Weibull. Error Structures: Independent errors and autoregressive AR(1) errors (code and outputs for the latter are located in the AC_ERROR directory). The simulations compare four estimators for the quantile ratio regression model: A) QRR1: Unregularized QRR with unit fixed effects. B) L2-QRR: Ridge-regularized QRR with empirical-Bayes tuning. C) QRR2: Cross-sectional QRR (implemented via the qrr function in the qtools R package). D) Naive QRR: Based on two separate fixed-effects quantile regressions. Repository Contents: Main R Scripts: Code to generate the data and execute all B = 500 replications for each baseline and AR(1) design. Workspaces: Saved .RData files containing the main simulation outputs. EBAYES_vs_CV.zip: Scripts and outputs for the sensitivity analysis comparing the empirical Bayes tuning rule against 5-fold cross-validation. README: Step-by-step instructions on how to rerun the simulations and reproduce the tables presented in the paper.