Reading data from csv file into time series with pandas

Question

My goal is to read EURUSD data (daily) into a time series object where I can easily slice-and-dice, aggregate, and resample the information based on irregular-ish time frames. This is most likely a simple answer. I'm working out of Python for Data Analysis but can't seem to bridge the gap.

After downloading and unzipping the data , I run the following code:

>>> import pandas as pd
>>> df = pd.read_csv('EURUSD_day.csv', parse_dates = {'Timestamp' : ['<DATE>', '<TIME>']}, index_col = 'Timestamp')

So far so good. I now have a nice data frame with Timestamps as the index.

However, the book implies (p. 295) that I should be able to subset the data, as follows, to look at all the data from the year 2001.

>>> df['2001']

But, that doesn't work.

Reading this question and answer tells me that I could import Timestamp:

>>> from pandas.lib import Timestamp
>>> s = df['<CLOSE>']

Which seems to work for a particular day:

>>> s[Timestamp('2001-01-04)]
0.9506999999

Yet, the following code yields a single value for my desired range of all data from year 2001.

>>> s[Timestamp('2001')]
0.8959

I know I am missing something simple, something basic. Can anyone help?

Thank you, Brian

Answer 1

The example on pg. 295 is being performed on Series object which is why indexing with the year works. With a DataFrame you would want df.ix['2001'] to achieve the same results.

Answer 2

If you want to get all of the columns, then df.ix['2001'] .

If you're interested only in "CLOSE", since you already did s = df['<CLOSE>'] , you can get the 2001 values by s['2001']