Minimum Information Copula

This repository contains Python codes used in our paper

On the minimum information checkerboard copulas under fixed Kendall's rank correlation
Issey Sukeda, Tomonari Sei
The University of Tokyo

Enjoy data analysis using minimum information copula models!

---

Setup

GitHub経由

pip install git+https://github.com/stardust-coder/minimum-information-copula

ローカルから

git clone https://github.com/stardust-coder/minimum-information-copula
pip install .

Usage

from MIC import *
import pandas as pd
import matplotlib.pyplot as plt

#Load dataset
df = dataloader.load_data_template("trading_jp", start=(2010,1,1),end=(2020,12,31),log_return=True) #currently only support 5 major Japanese trading firms.
df = df.iloc[:,[0,1]] #choose 2 columns.
###or
df = pd.read_csv("hoge.csv") #your own data

#Calculate statistics
tau,rho,_,_,_,_= stats.calc_bivar_stats(df)

#Fit 30x30 MICK/MICS to data. Other gridsizes will be coming soon...
#About 1min.
mick = copula.greedy_MICK(size=30,invariance=copula.plor(tau.statistic)) #need to convert tau into pseudo log odds ratio
mics = copula.greedy_MICS(size=30,invariance=copula.lor(rho.statistic)) #need to convert rho into log odds ratio

#Sample from copulas
samples = sampler.sample_from_ccopula(mick,sample_size=1000,save_csv=False)
plt.figure(figsize=(5,5))
plt.scatter(samples.values[:,0],samples.values[:,1])

#Visualize copula densities
visualize.three_dim_plot_flatten(mics, top=0.01) #top means the upper bound of the density

Run on App

streamlit run app.py

Document

Dataloader

Return Pandas DataFrame with 2 columns. dataloader.load_data_template("trading_jp", start=(2010,1,1),end=(2020,12,31),log_return=True)

Sampler

from Clayton copula sampler.MOsamples_clayton(alpha=10,size=100)
from Gumbel-Hougaard copula sampler.MOsamples_gh(gamma=5,size=100)
from Frank copula sampler.Invsamples_frank(theta=20,size=100)

Copula

To be done.

Stats

To be done.

Contact

If you have any comments, contact sukeda-issei006[at]g.ecc.u-tokyo.ac.jp