Coding Cross Correlation Function in Python

[DarthXander]

I have two data sets which I wish to perform the discrete correlation
function on and then plot the results for many values of t to see what
if any time lag exists between the data.
Thus far my code is;

import csv
import pylab
from pylab import *
from numpy import *
from numpy import array

HSBC=csv.reader(open("HSBC data.csv"))
Barclays=csv.reader(open("Barclays data.csv"))

x=[]
a=[]
y=[]
b=[]
g=[]
h=[]
d=[]

for Date, Close in HSBC:
x.append(Date)
a.append(float(Close))

for Date, Close in Barclays:
y.append(Date)
b.append(float(Close))

for index in range(len(a)):
g.append(a[index]-mean(a))

for index in range(len(b)):
h.append(b[index]-mean(b))

r=std(a)
s=std(b)

So I have all the necessary components for the DCF.

However I'm not faced with the challenge of performing the DCF for t
in the range of potentially 0-700 or so.
Currently I could do it individually for each value of tau ie;

t1=[]
for index in range(len(g)-1):
j=(g[index]*h[index+1])/(r*s)
t1.append(j)

d.append(mean(t1))

However to do this 700 times seems ridiculous. How would I get python
to perform this for me for t in a range of roughly 0-700?

Thanks
Alex
-- 
http://mail.python.org/mailman/listinfo/python-list

Re: Coding Cross Correlation Function in Python

[DarthXander]

On Dec 2, 7:12 pm, sturlamolden  wrote:
> For two 1D ndarrays, the cross-correlation is
>
> from numpy.fft import rfft, irfft
> from numpy import fliplr
>
> xcorr = lambda x,y : irfft(rfft(x)*rfft(fliplr(y)))
>
> Normalize as you wish, and preferably pad with zeros before invoking
> xcorr.

Thanks, though I'd like to do this longer hand than the built in
functions! Great to approximate it though.
-- 
http://mail.python.org/mailman/listinfo/python-list