Value-at-Risk for a portfolio

I have yet to see a comprehensive variance/co-variance method for getting VaR values for individual securities in a portfolio. So lets begin.

First, we need pandas, numpy, scipy and a datasource for our portfolio. I prefer 'bt' for the later.

import pandas as pd
import numpy as np
import pandas.io.data as web
import datetime
import matplotlib.pyplot as plt
import bt

from scipy.stats import norm

So lets pull the data

start = datetime.datetime(2013, 5, 1)


portfolio = bt.get('aapl, tlt, fxe, gs, c, fb, ms, ko, bac', start=start)
port = pd.DataFrame(index=portfolio.columns)

Now we have two dataframes. One with our historical values as portfolio, the other as port which holds the security names which we will later store our VaR values in. Next we create the actual VaRCoV function:

def var_cov_var(P, c, mu, sigma):
    alpha = norm.ppf(1-c, mu, sigma)
    return P - P*(alpha + 1)

def getVar(portfolio, c): #portfolio dataframe, confidence interval    P = 15000   #each position % of portfolio, should create a new one to store marketvalue of each position to be more precise    #c = 0.99    #confidence interval, moved to function    totalrisk = 0    for i in portfolio:
        rets = portfolio[i].pct_change()
        mu = np.mean(rets)
        sigma = np.std(rets)
        var = var_cov_var(P, c, mu, sigma)
        port["VaR"][i] = var
        totalrisk += var
    return port, totalrisk

getVar(portfolio, .95)

Out[471]: 
(             VaR
 aapl  328.585833
 tlt   205.453309
 fxe   136.978460
 gs    274.577792
 c     315.838626
 fb    566.777609
 ms    332.530446
 ko    220.500392
 bac   319.658819, 2700.9012842524335)

Chees.