How to do a series of series in pandas

Question

Say I have two series: a and b,

a = Series(None, index=['a','b','c'])
b = Series('lol', index=['j','k','l'])

I would like to store b as one of the elements of a,

a['a'] = b

but I get

ValueError: setting an array element with a sequence.

Is it possible to store a pandas series inside a pandas series? How can I do it? Thanks.

Series is a one-dimensional labeled array (technically a subclass of ndarray) capable of holding any data type (integers, strings, floating point numbers, Python objects, etc.). i think it's not possible. — Dmitry Zagorulkin, Jan 09 '13 at 12:53
Thanks @ZagorulkinDmitry, but isn't the pandas series a python object as well? This is the reason of my confusion. — Victor, Jan 09 '13 at 12:59

Andy Hayden · Accepted Answer · 2015-12-13T17:24:31.497

6

You can recast the dtype using the method astype:

In [11]: a = a.astype(object)

In [12]: a['a'] = b

In [13]: a
Out[13]: 
a    [lol, lol, lol]
b                NaN
c                NaN

Alternatively (to using astype) when contructing a you can force the dtype to be object:

In [14]: a = Series(None, index=['a','b','c'], dtype=object)

The reason you are getting this error is because float64, doesn't allow a Series and similarly it doesn't allow strings - try to set a['a'] = 'lol' and you'll get a ValueError.

In [21]: a = Series(None, index=['a','b','c'])

In [22]: a.dtype
Out[22]: dtype('float64')

You can read more about type-casting in the docs.

edited Dec 13 '15 at 17:24

answered Jan 09 '13 at 12:58

Andy Hayden

359,921
101
625
535

@ZagorulkinDmitry changed to use astype and link to the docs :) – Andy Hayden Jan 09 '13 at 13:09
Thank you Hayden! I'd buy you a pint if I were in London. – Victor Jan 09 '13 at 13:19

How to do a series of series in pandas

1 Answers1

Linked