Python pandas dataframe: interpolation using dataframe data without updating it. Just get the interpolated value.

Question

I am fairly new to Python pandas library and cannot find answer to my problem in other posts. I have a dataframe that looks like this. Dates are index names and series are column names.

>>> MyDataframe
             Serie1  Serie2  Serie3  Serie4  Serie5 
2011-04-30      92      96     NaN     NaN     NaN  
2011-05-31     164     168      12     16      NaN
2011-06-30     238     242      90     20      88
2011-07-31     322     326     169     120     167

I would like to perform 1D linear interpolations within this dataframe but without modifying the dataframe, I just want to get the result. For instance I want to determine what is the value of Serie2 at the date of 2011-06-10. The functions DataFrame.interpolate() and Series.interpolate() seem to be useful only to replace the NaN with interpolated data.

Is there a function that could perform something like:

Result = MyDataFrame['Serie2'].interpolate('2011-06-10')

and it would simply return the linear interpolation between 168 and 242.

Thanks in advance for your support!

Answer 1

interpolate interpolates using the existing index, so you have to reindex the df and then call interpolate :

In [48]:
df.reindex(pd.date_range(df.index[0], df.index[-1])).interpolate().loc['2011-06-10']

Out[48]:
Serie1    188.666667
Serie2    192.666667
Serie3     38.000000
Serie4     17.333333
Serie5           NaN
Name: 2011-06-10 00:00:00, dtype: float64

Once this is done you can select a specific date and column:

In [49]:
df.reindex(pd.date_range(df.index[0], df.index[-1])).interpolate().loc['2011-06-10']['Serie2']

Out[49]:
192.66666666666666

Here I generate a new datetimeindex using the first and last values in your index using date_range .

It will be more efficient to just interpolate between the existing index values that are in your range.

We can find the lower bound of the index value using get_slice_bound :

In [70]:
start = df.index.get_slice_bound('2011-06-10', side='right', kind=None)

df.reindex(pd.date_range(df.index[start-1], df.index[start])).interpolate().loc['2011-06-10']['Serie2']
Out[70]:
192.66666666666666