Converting PeriodIndex to DateTimeIndex?

I have a question regarding converting a tseries.period.PeriodIndex into a datetime.

I have a DataFrame which looks like this:

               colors      country
time_month 

2010-09         xxx        xxx
2010-10         xxx        xxx
2010-11         xxx        xxx
...

time_month is the index.

type(df.index)

returns

class 'pandas.tseries.period.PeriodIndex'

When I try to use the df for a VAR analysis (http://statsmodels.sourceforge.net/devel/vector_ar.html#vector-autoregressions-tsa-vector-ar),

VAR(mdata)

returns:

Given a pandas object and the index does not contain dates

So apparently, Period is not recognized as a datetime. Now, my question is how to convert the index (time_month) into a datetime the VAR analysis can work with?

df.index = pandas.DatetimeIndex(df.index)

returns

cannot convert Int64Index->DatetimeIndex

Thank for your help!

joris

You can use the to_timestamp method of PeriodIndex for this:

In [25]: pidx = pd.period_range('2012-01-01', periods=10)

In [26]: pidx
Out[26]:
<class 'pandas.tseries.period.PeriodIndex'>
[2012-01-01, ..., 2012-01-10]
Length: 10, Freq: D

In [27]: pidx.to_timestamp()
Out[27]:
<class 'pandas.tseries.index.DatetimeIndex'>
[2012-01-01, ..., 2012-01-10]
Length: 10, Freq: D, Timezone: None

In older versions of Pandas the method was to_datetime

You can also use the following to get exactly the same result.