Malte Knueppel and Fabian Krueger, "Disagreement, Uncertainty and the
Linear Pool", Journal of Applied Econometrics, Vol. 37, No. 1, 2022,
pp. 23-41.

Al files are zipped in the file kk-files.zip. It is quite large,
mainly because of the files in the directory ucsv_forecasts/ . The zip
contains both ASCII (text) files and binary files. The text files are
in Unix format. If Windows users encounter difficulties with any of
these files, they may need to convert them to DOS format.

-----------------------------------------------------------
Codes for Monte Carlo Simulations (Section 5 of the paper)
-----------------------------------------------------------

-- The R script "Section5_MC_Computations.R" produces the simulation
results for two designs reported in the paper (see variable "design"
defined in Line 32). The output file is called "MC_results_[d].RData",
where [d] describes the design used.

-- The R script "Section5_MC_Figures.R" produces Figures 1 and 2 in
the paper, based on the simulation results from the previous step. The
figures are saved in a subfolder called "outputs/". 


-----------------------------------------------------------------
Code and data for the Empirical Analysis (Section 6 of the paper)
-----------------------------------------------------------------

-- The Excel file "PQvQd.xlsx" contains real time vintage data on the
GDP deflator. They were obtained from the Federal Reserve Bank of
Philadelphia's real time data set at
https://www.philadelphiafed.org/surveys-and-data/real-time-data-research/p

-- The Excel file "Individual_PGDP.xlsx" contains inflation forecasts
from the Survey of Professional Forecasters run by the Federal Reserve
Bank of Philadelphia. The data were obtained from their website at
https://www.philadelphiafed.org/surveys-and-data/pgdp. 

-- Disclaimer: The Federal Reserve Bank of Philadelphia does not take
responsibility for the accuracy of the data posted here. Furthermore,
the raw data may be revised over time, and the websites linked above
should be consulted for the official, most recent versions.

-- The Matlab script "main_UCSV_rt.m" produces MCMC forecast draws
from the UCSV model using the Matlab script "UCSV_noMA.m" and
functions in the subfolder "matlab_functions". The scripts are based
on the codes generously provided by Joshua Chan at
https://joshuachan.org/code/MASV_matlab.zip, for details see the
headers of the scripts.  The output of "main_UCSV_rt.m" is stored in a
subfolder called "ucsv_forecasts". It consists of ten .csv files: The
files called "mean_y_fore_all_h1.csv" to "mean_y_fore_all_h5.csv"
contain mean forecasts at horizons 1 to 5; the files called
"vari_y_fore_all_h1.csv" to "vari_y_fore_all_h5.csv" contain variance
forecasts at horizons 1 to 5. NOTE: Since the .csv files are already
provided, you can run the following R scripts without having to run
the Matlab script first. The .csv files provided are those used for
the paper.

-- The R script "Import_UCSV_forecasts.R" loads the .csv files
produced in the previous step and saves them as a single .RData file
that can easily be loaded into R (see below). 

-- The R script "procs_rev.R" contains various functions, and is
imported by some of the other files mentioned below.

-- The file "bvarsv3_0.01.tar.gz" is an R package (written in R and
C++) for estimating the CMM model considered in Section 6. To install
the package from within R, please set the working directory to the
folder containing the ".tar.gz" file, and enter
install.packages("bvarsv3_0.01.tar.gz", repos = NULL, type = "source")
Note that the package is not fully documented. It is partly based on
Elmar Mertens' Matlab code at
https://github.com/elmarmertens/CMMrestat-TimeVaryingUncertainty. We
thank Elmar Mertens for sharing his code, and Justus Thomsen for
excellent research assistance in translating the code to R and C++. 

-- The R script "Section6_Empirics_Computations.R" fits the CMM model
and several forecast combination models. It produces a file called
"results_[date].RData". Note that running the file takes several
hours. One result of such a run ("results_2021-03-31.RData") is
included in the folder.

-- The R script "Section6_Empirics_Figures_Tables.R" produces Tables
2-3 and Figures 3-5 in the paper, based on the results from the
previous step. The figures and tables are saved in a subfolder called
"outputs/". 


Contact: fabian.krueger [AT] kit.edu