""" Created on Thursday March 19 @author: Jeffrey J. Walker test_reader.py Simple script to test the yahoo_csv_reader.py function. """ import pandas as pd import numpy as np import matplotlib.pyplot as plt ## insert the path corresponding to the Yahoo csv reader. # insert at 1, 0 is the script path (or '' in REPL) sys.path.insert(1, '/home/jjwalker/Desktop/finance/codes/data_cleaning') from yahoo_csv_reader import yahoo_csv_reader ## path to the security you want to plot: path = '/home/jjwalker/Desktop/finance/data/commodities/' ticker = 'HG=F' filename = path + ticker df = yahoo_csv_reader(filename, ticker) ## What is a good default plot, to include the chart, and maybe the volume? plt.plot(df.Close, '.k') plt.show()
from statsmodels.tsa.stattools import grangercausalitytests
from statsmodels.tsa.stattools import adfuller
#from statsmodels.tsa.stattools import kpss
## import the csv reader for Yahoo finance data
# insert at 1, 0 is the script path (or '' in REPL)
sys.path.insert(1, '/home/jjwalker/Desktop/finance/codes/data_cleaning')
from yahoo_csv_reader import yahoo_csv_reader
## import the csv reader for FRED data
from fred_csv_reader import fred_csv_reader
## path where the various data files are located.
spx_path = '/home/jjwalker/Desktop/finance/data/stocks/'
fed_path = '/home/jjwalker/Desktop/finance/data/us_economic_data/'
fed_bs = fred_csv_reader(fed_path + 'WALCL')
spx = yahoo_csv_reader(spx_path + '^GSPC', '^GSPC')
## Want to execute in python shell? then use:
#execfile('/home/jjwalker/Desktop/finance/codes/stocks/granger_causality.py')
## concatenate the close price for the two (or more) securities and find the correlation.
common_values = pd.concat([fed_bs.WALCL, spx.Close], axis=1)
## get rid of any nans?? Or, resample based on WALCL?
common_values.dropna(inplace=True)
## choose a startdate?
start_date = pd.to_datetime('2009-01-02')
## calculate pearson correlation coefficient, easy peasy:
corr_coef = common_values[common_values.index > start_date].corr(
method='pearson')
## BEFORE RUNNING ADF TEST, YOU NEED TO DO KPSS TEST!
from statsmodels.tsa.stattools import adfuller
##~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
## Be sure that you use a valid ticker symbol, no indices!
## This will be the stock you are long on.
ticker = '^SPX'
## insert the path corresponding to the Yahoo option chain scraper;
## we will need this function!
# insert at 1, 0 is the script path (or '' in REPL)
sys.path.insert(1, '/home/jjwalker/Desktop/finance/codes/data_cleaning')
from yahoo_csv_reader import yahoo_csv_reader
## the path:
path = '/home/jjwalker/Desktop/finance/data/stocks/'
df = yahoo_csv_reader(path + ticker, ticker)
## ADF test: Null hypothesis is that the time series has unit root
adftest = adfuller(df.Close, autolag='AIC')
dfResults = pd.Series(adftest[0:4],
index=[
'ADF Test Statistic', 'P-Value', '# Lags Used',
'# Observations Used'
])
## critical values come from
#adfResults[4].items()
## p-values from adfResults[1]
## The more negative the test statistic, the more likely we are to reject the
## NULL HYPOTHESIS that there is unit root.
## If the test statistic is less than the critical value, we reject Null
## hypothesis
'DGS20': '20',
'DGS30': '30',
},
inplace=True)
## restructure the dataframe, sort the columns numerically
## However, this sorts alphabetically instead of numerically
#yc.sort_index(axis=1,ascending=True,inplace=True)
## FINALLY found the way to do this:
yc = yc[np.argsort(yc.columns.astype(float))]
## standard us treasury maturities
xtemp = np.array(yc.columns.astype(float))
## Load spx for comparison: (or use ^GSPC)
## EVEN BETTER! Use sp500TR, the sp500 total return index which assumes
## dividend reinvestment!
spx = yahoo_csv_reader('/home/jjwalker/Desktop/finance/data/stocks/^SP500TR',
'^SP500TR')
## SPX daily return, in case we need it:
spx_drf = spx.Close[1:].values / (spx.Close[:-1])
## Zero coupon yield curve from FRED:
## read in data; have to use (7)th line as the header, I don't know why
datafile = "/home/jjwalker/Desktop/finance/data/bonds/feds200628.csv"
zc = pd.read_csv(datafile, header=7)
zc['Date'] = pd.to_datetime(zc.Date, infer_datetime_format=True)
zc.set_index('Date', inplace=True)
## the SVENY variable names contain the zero-coupon yield, continuously
## compounded
## get the variables needed:
## 1 year zero coupon (for annual Sortino/information/Sharpe ratio)
zc01 = zc['SVENY01']
## 30 year zero coupon, the instrument we are interested in
'DGS30': '30',
},
inplace=True)
## restructure the dataframe, sort the columns numerically
## However, this sorts alphabetically instead of numerically
#yc.sort_index(axis=1,ascending=True,inplace=True)
## FINALLY found the way to do this:
yc = yc[np.argsort(yc.columns.astype(float))]
## standard us treasury maturities
xtemp = np.array(yc.columns.astype(float))
## data for spx: (or use ^GSPC)
## EVEN BETTER! Use sp500TR, the sp500 total return index which assumes
## dividend reinvestment!
#spx=yahoo_csv_reader('/home/jjwalker/Desktop/finance/data/stocks/^SP500TR','^SP500TR')
spx = yahoo_csv_reader('/home/jjwalker/Desktop/finance/data/stocks/^GSPC',
'^GSPC')
## SPX daily return, in case we need it:
#spx_drf=spx.Close[1:].values/(spx.Close[:-1])
## DCA inputs:
## starting cash; make this zero if you want to start by dca'ing first
## iteration of list
cash = 0
## DCA investing amount, dollars
dca_inv = 200
## initial number of spx shares:
nshares = 0
## Dates to use for dollar cost averaging; set to the first of the month
## Find a start date; Ideally this would be a treasury auction date?
#start_date='1985-12-02' # this start date is the first month that 30 year
# STRIPS data is available
import matplotlib.pyplot as plt ## Need this line to import the yahoo csv reader # insert at 1, 0 is the script path (or '' in REPL) sys.path.insert(1, '/home/jjwalker/Desktop/finance/codes/data_cleaning') from yahoo_csv_reader import yahoo_csv_reader ## path where the various data files are located. path='/home/jjwalker/Desktop/finance/data/stocks/' ## load data for each ticker ticker1='NSC' #ticker2='AMSC' div_data=ticker+'_div' ## Start date? stock=yahoo_csv_reader(path+ticker1,ticker1) div=yahoo_csv_reader(path+div_data,ticker1) ## make sure we have sorted by index! div.sort(inplace=True) ## initial number of shares or initial investment. ## initial investment p0=1e4 ## initial shares: n0=np.floor(p0/stock.Close[stock.index==div.index[0]])#n0=10.0 ## number of shares grows with time, can be fractional. ## just use the paid date for when these get added to the principal. #start_date= n=np.zeros((len(div)) n[0]=n0 for i in range(len(div)-1):