CODEWITHSUNDEEP
pythonpandasdataframetime-seriesresampling
Nyxynyx
Nyxynyx

Reputation: 63609

Pandas Resampling error: Only valid with DatetimeIndex, TimedeltaIndex or PeriodIndex

When using pandas' resample function on a DataFrame in order to convert tick data to OHLCV, a resampling error is encountered.

How should we solve the error?

enter image description here

# Resample data into 30min bins
bars = data.Price.resample('30min', how='ohlc')
volumes = data.Volume.resample('30min', how='sum')

This gives the error:

TypeError: Only valid with DatetimeIndex, TimedeltaIndex or PeriodIndex, but got an instance of 'Int64Index'

Upvotes: 57

Views: 102455

Answers (2)

cottontail
cottontail

Reputation: 23041

Because it is designed for time-series data, as the error says, resample() works only if the index is datetime, timedelta or period. The following are a few common ways this error may show up.

However, you can also use the on= parameter to use a column as grouper, without having a datetime index.

df['Timestamp'] = pd.to_datetime(df['Timestamp'], unit='s')
bars = df.resample('30min', on='Timestamp')['Price'].ohlc()
volumes = df.resample('30min', on='Timestamp')['Volume'].sum()

res1


If you have a MultiIndex dataframe where one of the index is datetime, then you can use level= to select that level as the grouper.

volumes = df.resample('30min', level='Timestamp')['Volume'].sum()

res2


You can also use resample.agg to pass multiple methods.

resampled = df.resample('30min', on='Timestamp').agg({'Price': 'ohlc', 'Volume': 'sum'})

res3

Upvotes: 2

unutbu
unutbu

Reputation: 879163

Convert the integer timestamps in the index to a DatetimeIndex:

data.index = pd.to_datetime(data.index, unit='s')

This interprets the integers as seconds since the Epoch.

For example, given

data = pd.DataFrame(
    {'Timestamp':[1313331280, 1313334917, 1313334917, 1313340309, 1313340309], 
     'Price': [10.4]*3 + [10.5]*2, 'Volume': [0.779, 0.101, 0.316, 0.150, 1.8]})
data = data.set_index(['Timestamp'])
#             Price  Volume
# Timestamp                
# 1313331280   10.4   0.779
# 1313334917   10.4   0.101
# 1313334917   10.4   0.316
# 1313340309   10.5   0.150
# 1313340309   10.5   1.800

data.index = pd.to_datetime(data.index, unit='s')

yields

                     Price  Volume
2011-08-14 14:14:40   10.4   0.779
2011-08-14 15:15:17   10.4   0.101
2011-08-14 15:15:17   10.4   0.316
2011-08-14 16:45:09   10.5   0.150
2011-08-14 16:45:09   10.5   1.800

Then

ticks = data.ix[:, ['Price', 'Volume']]
bars = ticks.Price.resample('30min').ohlc()
volumes = ticks.Volume.resample('30min').sum()

can be computed:

In [368]: bars
Out[368]: 
                     open  high   low  close
2011-08-14 14:00:00  10.4  10.4  10.4   10.4
2011-08-14 14:30:00   NaN   NaN   NaN    NaN
2011-08-14 15:00:00  10.4  10.4  10.4   10.4
2011-08-14 15:30:00   NaN   NaN   NaN    NaN
2011-08-14 16:00:00   NaN   NaN   NaN    NaN
2011-08-14 16:30:00  10.5  10.5  10.5   10.5

In [369]: volumes
Out[369]: 
2011-08-14 14:00:00    0.779
2011-08-14 14:30:00      NaN
2011-08-14 15:00:00    0.417
2011-08-14 15:30:00      NaN
2011-08-14 16:00:00      NaN
2011-08-14 16:30:00    1.950
Freq: 30T, Name: Volume, dtype: float64

Upvotes: 74

Related Questions