George Kapetanios, Massimiliano Marcellino, and Fabrizio Venditti,
"Large Time-Varying Parameter VARs: A Non-Parametric Approach",
Journal of Applied Econometrics, Vol. 34, No. 7, 2019, pp. 1027-1049.

All files are in kmv-files.zip. They are a mix of ASCII and binary
files, and there are many folders and subfolders.

**** Data sources

The spreadsheet largedataset_1.xlsx contains the raw data with a
description of each series.

All data are from FRED at https://research.stlouisfed.org/

The code Master_prepare_data.m transforms the raw data when needed
(logs or differences) and generates the file data_20_78.mat that is
used throughout the empirical analysis. 

To replicate the results of the paper see the description below. 

***** Replicating Forecasting Exercises.

************ Out of Sample Forecasting

Run the code Master_Forecasting_JAE.m selecting mods either 1 or 2 
(mods = 1 for 20 variables model, 2 for 78 variables model).

This runs the out of sample forecasting exercise and stores the
results in the folder /output_forecast.

Run the code EvaluateForecast_JAE. This produces Tables 3 to 5 in the
paper and Charts 1 to 3.


************ Structural Analysis

The code Master_structural_JAE does structural analysis and the charts
of the IRFs contained int the paper.

************ MONTE CARLO

The code MasterJAE_MCARLO produces the Monte Carlo Estimates. To
produce the tables that appear in the paper, run the code
MasterJAE_tables_Mcarlo .

************ Computational Gains exercise

MasterJAE_ComputationalGains estimates the computational gains of
parametric and non parametric estimators that are discussed in the
paper's Appendix.