I'm trying to do a copula GARCH model with data from 5 indexes (SP500, STXE600, Nikkei, FTS100 and PHLXGOLD). My data is already in a dataframe and I modeled the 4 GARCH model I will work with:
- GARCH-n (GARCH model with normal distribution): results
- GARCH-t (GARCH model with t student distribution): tresults
- GJR-n (Glosten, Jagannathan, and Runkle model with normal distribution): res
- GJR-t (Glosten, Jagannathan, and Runkle model with t student distribution): tres
After that I get the residuals for each model:
- GARCH-n: residuals_df
- GARCH-t: tresiduals_df
- GJR-n: res_df
- GJR-t: tres_df (this dataframes have different values, I checked)
Now I have to compute Gaussian, t-student, Frank, Clayton copulas for this data. I tried the following python libraries:
- Copulae: when I did (for example) Gaussian GARCH-n result I get the same values for Gaussian GARCH-t
- pycop: doen't have multivariate copulas automatic computaded
- Copulas : same as Copulae
My goal is to get the parameters of each copula and compute the LLF (log likelihood function), AIC and BIC to evaluate the best model. Does anyone ever had this problem before, and if so how did you solve it? Thanks