Eduardo Rossi and Paolo Santucci de Magistris, "Indirect Inference
with Time Series Observed with Errors", Journal of Applied
Econometrics, Vol. 33, No. 6, 2018, pp. 874-897.
All files are stored in the file rm-files.zip.
Data:
Data are stored both in Matlab ("Data.mat") and ASCII ("Data.txt",
comma separated) formats. Each file contains two series of 1008
observations on daily realized variance of JPMorgan computed both at 5
seconds and 5 minutes. Data files are in the folder "SV Estimation".
There are two folders which contain a number of MATLAB functions used
to generate the numerical and empirical results shown in the paper and
in the Supplementary material.
Codes:
The folder "OU model" contains the MATLAB codes to reproduce the
results of Table 1 in the Supplementary material. The code "Main OU"
performs the Monte Carlo simulation of the Ornstein-Uhlenbeck process
plus noise. A number of MATLAB functions are called to perform the
indirect inference estimation. In particular,
a) simulate_OU_exact.m: uses exact discretization to generate a
trajectory for the OU model
b) crude_discretrization_OLS.m: Adopts a crude discretization (AR1) to
estimate the auxiliary coefficients
c) indirect_inference_OU_crude_no_noise.m: Computes the criterion
function for the II estimation when the noise is neglected
d) indirect_inference_OU_crude_noise.m: Computes the criterion
function for the II estimation when the noise is included among the
structural features
The folder "SV Estimation" contains the MATLAB codes and the RV data
to reproduce the results of Table 2 in the article. In particular we
consider the case of a contamination of the log-returns with a noise
following an MA(1) process (the case with iid noise can be obtained
restricting the MA parameter to 0).
The main file "Main_MA_Noise_TF_Heston.m" loads the RV data for JPM
contained in the file "Data.mat". The main file calls a number of
functions:
a) "HAR.m": estimates the HAR model of Corsi.
b) "II_2FHeston_MA_noise.m": produces the criterion function used for
the II estimation based on the two-factors Heston model contaminated
by MA noise
c) "simulate_heston_two_factor.m": generates trajectories for the
two-factors Heston model
d) "Binding_2FHeston_MA_noise.m": Approximates the binding function of
the two-factors Heston model contaminated by MA noise. This is used to
compute the Jacobian.