This is the readme.txt file to replicate the results in 
the technology-Bayesian-VAR paper.

Each zip file contains all the files needed to replicate the 
results for one prior only. So you must have six zip files in total.

-----------------------------------------------------------
FILES DESCRIPTION

The data file: CEVdata.dat
ALL impulse responses in an Excel file: all-impulse-Bayesian-VAR.xls
Command file to follow the var.exe on the DOS prompt: *.CMD
The compiled Kadiyala-Karlsson FORTRAN: var.exe
To compute the posterior odds: posteriorodds.m

The var.exe is a binary DOS executable file
It is the compiled 'Dr. Sune Karlsson' executable 
with some modifications made. If you want to compile the FORTRAN 
source on your machine, please download the source files 
from their web site.

--------------------------------------------------------------
STEP BY STEP

(In the following replace '*' by: diff, min, en, enc, nd, nw)

For each prior:

1) Find out the values for the hyperparameters
go to the chooseprior directory, and then run at the DOS prompt
'var chooseprior*.CMD'  (replace * by the prior name: min, en, enc, nd, nw)
this will give you the 'chooseprior*.out' file. This file contains
the results. Edit it using any text editor to get the grid.
Copy the grid into the 'chooseprior*.csv' file.
Now, using MATLAB, run the 'chooseprior*.m' This will give you the 
'chooseprior_matlab*.out' file which tells you where the minimum is.


for example: TRY THIS IN THE 'EN' DIRECTORY
for the extended natural conjugate unconditional prior
at the DOS prompt run: var choosepriorEN.CMD
(note that this might take a day or two runnning
depending on how fast your computer is). The results are generated
in the 'choosepriorEN.out'.
Text edit 'choosepriorEN.out' and save it as 'choosepriorENdata.out'
copy the 'choosepriorENdata.out' to 'choosepriorENdata.csv'
now, using MATLAB, run the 'choosepriorEN.m'
This will generate the 'chooseprior_matlabEN.out' file 
needed for TABLE 2 in the paper
repeat the above for every prior.

The only prior that has a graph is the 'Minnesota' prior
which is in the paper (Figure 3). If you want to produce
more graphs for each prior, then use the 'chooseprior.m'
in the 'min' directory and copy it to other directories (priors)
and then run it in MATLAB.

2) Once you know which values to take for the hyperparameters, 
go to the main directory for the prior and plug in the values in the
'LITT*.CMD.txt' then save it as (overwrite) 'LITT*.CMD'

3) at the DOS prompt, run: 'var LITT*.CMD'

4) This will generate the posterior for all parameters in the 
'LITT*.txt' file and at the same time it will generate 
the 'LITT*.out' file which contains the results of the VAR
(parameters, var-cov, etc.)

5) In MATLAB, run the 'litt*.m' file. This will read the 'LITT*.txt'
will build the VAR, impose the identifications: Blanchard-Quah and 
sign restrictions, and then produce the impulse responses and its 
standard errors, as well as the posterior histogram for 
each parameter in the VAR. Save the graphs under the 
directory 'figures' 

6) the long-run multipliers and the convergence diagnosis results are
in the '*diary.out'

7) the figures are already saved for you in the 'figures' directory.

7) to compute the posterior odds, use 'posteriorodds.m'

------------------------------------------------------------------------


O. Mikhail
http://www.bus.ucf.edu/omikhail
omikhail@bus.ucf.edu
omikhail@hotmail.com