bashtage/mfe-toolbox

mfe-toolbox's Issues

Output issue: Calling figarch from Python 3.5

Hello everyone,

I am using matlab.engine to call figarch from python 3.5 and I have noticed an issue. If I run figarch from matlab using figarch([0.01 0.01 0.01],1,1), my output will return the parameters including all other outputs such as the Logarithmic Likelihood. However, if I run the exact same parameters for figarch([0.01 0.01 0.01],1,1) using python, the program will only return the parameters and nothing else.

---res.resid is just the residuals from a previous process---

import matlab.engine 
eng = matlab.engine.start_matlab()
eng.cd(path.expanduser('~/Documents/MATLAB/MFEToolbox/univariate'))
print(eng.figarch(eng.transpose(matlab.double(res.resid.T.values.tolist())), 1, 1, 'SKEWT'))

In addition, if I try to include starting values for the parameters used in figarch, there is a small bug in the figarch_starting_values.m for the case of non-empty startingvals (where an additional parameter 'o' is undefined).

I am sure that figarch should return all outputs as indicated in the documentation in either Matlab or calling from Python, I don't know what I missed.

Would really really appreciate if someone can shed some light!!!

Thank you!

Sincerely,
Terence

Output issue on Figarch.m file

Hello everyone,
I'm using the functions on MFE Toolbox in order to estimate a FIGARCH. There is a problem in line 223: the command 'robustvcv' no longer exists, so if you compute the function figarch with more than 3 nargouts returns an error. The ouputs VCV, robustVCV are necessary to infer the model and assess the robustness of the estimation.

if nargout>3
nw=0; %No newey west on scores
[VCVrobust,A,B,scores,hess]=robustvcv('figarch_likelihood',parameters,nw,
p,q,epsilon,epsilon2Augmented,truncLag,errorType);
T = size(epsilon,1);
VCV=hess^(-1)/T;
end

Would really really appreciate if someone can help me.

Thank you!

Sincerely,
Alessandro