StarVine

The StarVine package is split into four subpackages:

• vine : C- and R- vine classes
• bvcopula : Bivariate copula
• mvar : Multi-variate data
• uvar : Uni-variate data (marginal distribution fitting)

Quick Example

Example return data is provided in the /tests/data/stocks.csv file. The following example will fit a C-vine to a 5 dimensional data set.

Setup and load the data file:

from starvine.vine.C_vine import Cvine
import pandas as pd
stocks = np.loadtxt(dataDir + 'stocks.csv', delimiter=',')
ddict = {
    '1a': stocks[:, 0]
    '2b' : stocks[:, 1]
    '3c' : stocks[:, 4]
    '4d' : stocks[:, 5]
    '5e' : stocks[:, 6]
}
tstData = pd.dataFrame(ddict)

Visualize raw data:

from starvine.mvar.mv_plot import matrixPairPlot
matrixPairPlot(tstData, savefig="quad_varaite_ex.png")

Raw Data

Rank data and visualize:

ranked_data = tstData.dropna().rank()/(len(tstData)+1)
matrixPairPlot(ranked_data, savefig="quad_varaite_ranked_ex.png")

Ranked Data

Fit and plot the vine.

tstVine = Cvine(tstData)
tstVine.constructVine()
tstVine.plotVine(savefig="c_vine_graph_ex.png")

Visualize the C-vine structure:

C-vine

Sample the vine:

samples = tstVine.sample(n=8000)
matrixPairPlot(samples, savefig="quad_varaite_resampled_ex.png")

Resampled Data