Loaiza-Maya, R., Smith, M. S., and Maneesoonthorn, W. (2017). "Time
Series Copulas for Heteroskedastic Data", Journal of Applied
Econometrics, Vol. 33, No. 3, 2018, pp. 332-354.

The zip file lsm-files.zip contains a set of MATLAB routines used for
estimating the univariate exchange rate returns aplication in section
2.3 the paper. This code can be used for academic purposes as long as
the paper is cited.

The zip file also contains the data in Data.csv. There are 3670
observations on three exchange rates.

Below you can find a brief description for each routine.

/*------------------------------  Main routine -------------------------*/
<Main_DVineCop.m>
This is the main script. Run as it is, it estimates the posterior
distribution of the parameters of the DVine copula with mixture of
generalized Gumbel copula components applied to the USD/AUD data
returns. A Markov order of 1 is set by default.

/*---------------------------  Important routines ----------------------*/
<set_constants.m>
This routine provides options for running the model using different
exchange rates, Markov orders and copula families.  Copulas A1, A5,
B1, B5, T1, and T5 in the validation section 2.3.2 can be estimated
following the intructions therein.

<set_initial_states.m>
Initial values for the parameters and Random Walk Metropolis Hastings steps.

<set_priors.m>
Creates the parameter priors.

<gen_gamma.m>
Generates pair-copula parameters using RWMH.

<step_update.m>
Adapts steps for RWMH so the acceptance rates stay between 15% and 30%.

/*------------------  Remaining routines -------------------------------*/
The routines <lnc_gen_gumbel,lnc_gen_gum_mix,lnc_mix,lnc_DVine> are 
all used to compute the logarithm of c^cG, c^mix and c^DV in the paper.

The routines
<h_bb7,h_gumbel,h_gen_gumbel,h_gen_gumbel_mix,h_tcop,h_tcop_mix,h_mix_cop>
are all used to compute h functions for different copula families. All
of them are encompassed by the more general routine <h_func>; set its
third agument to the copula of choice.

The routines <fit_ssvkernel,ssvkernel> are used for fitting an
adaptive kernel density to the returns data.

The routines <expma,U_transf> are used for computational stability.

The routine <Truncatednormrnd> generates from a truncated normal
distribution.