Lei Jiang, Esfandiar Maasoumi, Jiening Pan, and Ke Wu, "A Test of
General Asymmetric Dependence", Journal of Applied Econometrics,
Vol. 33, No. 7, 2018, pp. 1026-1043.

All files are zipped in jmpw-files.zip, which has five folders within
it. ASCII files are in DOS format.

All source/function codes are written in R (with embeded comments),
and included in the folder named "FuncLib". These are called by other
R programs in our simulation/empirical analysis.

-- KLdiv.R is used to compute the sample estimates of the
Kullback-Leibler exceedance mutual information measure on tails.

-- npcdepKL.R is the code for stationary bootstrap.

-- exceed_cor.R and exceed_corSBpiv.R are used to compute exceedance
correlation and conduct the Hong, Tu, Zhou (2007) test via stationary
nested bootstrap method, respectively.

-- exceed_correl.R is the R version of Andrew Patton's Matlab code for
conducting the original Hong, Tu, Zhou (2007) test via asymptotic
distribution.


The main properties of our MI-based asymmetric dependence test can be
reproduced using programs and training data contained in the folder
named "simulation".

-- Subfolder "simDGPs" contains all the programs used to simulate the
return data via Copula-GARCH processes for all the benchmark case and
other variant cases.

-- Subfolder "comp_pvals" contains programs used to compute the size
and power of the MI-based test and the HTZ test. All the reported
results in Tables 1, 2, 3, 4, 5, and 7 can be reproduced with those
programs or with slightly modified versions of them. We rely on some
codes from Geir Berentsen (Geir.Berentsen@uib.no) to estimate local
Gaussian correlations and produce the results in Table 3. We refer the
interested readers to request those codes directly from him. 

The empirical part of the paper uses three datasets--the first two are
portfolio return datasets downloaded from Kenneth French's website:

  http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html

The dataset "data/US/US.csv" is returns of portfolios that are
constructed by sorting stocks based on their lagged size (sz),
book-to-market (bm) and momentum (mom) in the US market.

The data set "Canada.csv", "France.csv", "Germany.csv", "Japan.csv",
"Swtzrlnd.csv", and "UK.csv" in the folder "data/InterData" are
returns of portfolios sorted using lagged bm, earnings to price (ep)
and cash earnings to price (cep) in each market listed above. The
third dataset "combined.csv" is the MSCI stock index for the G7
developed countries, which can be found in the folder
"data/InterData/index".

Files in the folder "code" are programs we used to conduct the
empirical analyses (Table 8) in the paper.

Jiening Pan
School of Finance
Nankai University

jiening.pan [AT] nankai.edu.cn