Question

When using pandas stats.moments.rolling_mean(array,window) function I noticed that putting an extra argument changes the output, and is only padded with nans in the beginning not the end.

In[1]: import pandas as pd

In[2]: pd.stats.moments.rolling_mean(np.arange(12),6)
Out[2]: 
array([ nan,  nan,  nan,  nan,  nan,  2.5,  3.5,  4.5,  5.5,  6.5,  7.5,
        8.5])

I expected there to be 6 nans: 3 at the beginning and 3 at the end.
What am I missing here?

/M

Answer 1

The pandas implementation uses a rolling window of the previous n values, which is how it’s usually done in finance (see this Wikipedia entry for simple moving average).

I guess it would be nice to have the option to specify whether the values should be taken from either side or just use previous values – you can raise an issue on GitHub.

len(np.arange(12)) and len(pd.stats.moments.rolling_mean(np.arange(12),6)) both equal 12 as I would have expected – what result were you expecting?