I have a dictionary with each key holding a list of float values. These lists are not of same size.
I'd like to convert this dictionary to a pandas dataframe so that I can perform some analysis functions on the data easily such as (min, max, average, standard deviation, more).
My dictionary looks like this:
{
'key1': [10, 100.1, 0.98, 1.2],
'key2': [72.5],
'key3': [1, 5.2, 71.2, 9, 10.11, 12.21, 65, 7]
}
What is the best way to get this into a dataframe so that I can utilize basic functions like sum, mean, describe, std?
The examples I find (like the link above), all assume each of the keys have the same number of values in the list.
d={
'key1': [10, 100.1, 0.98, 1.2],
'key2': [72.5],
'key3': [1, 5.2, 71.2, 9, 10.11, 12.21, 65, 7]
}
df=pd.DataFrame.from_dict(d,orient='index').transpose()
Then df is
key3 key2 key1
0 1.00 72.5 10.00
1 5.20 NaN 100.10
2 71.20 NaN 0.98
3 9.00 NaN 1.20
4 10.11 NaN NaN
Note that numpy has some built in functions that can do calculations ignoring NaN values, which may be relevant here. For example, if you want to find the mean of 'key1' column, you can do it as follows:
import numpy as np
np.nanmean(df[['key1']])
28.07
Other useful functions include numpy.nanstd, numpy.nanvar, numpy.nanmedian, numpy.nansum.
EDIT: Note that the functions from your basic functions link can also handle nan values. However, their estimators may be different from those of numpy. For example, they calculate the unbiased estimator of sample variance, while the numpy version calculates the "usual" estimator of sample variance.
your_dict = {
'key1': [10, 100.1, 0.98, 1.2],
'key2': [72.5],
'key3': [1, 5.2, 71.2, 9, 10.11, 12.21, 65, 7]
}
pd.concat({k: pd.Series(v) for k, v in your_dict.items()})
key1 0 10.00
1 100.10
2 0.98
3 1.20
key2 0 72.50
key3 0 1.00
1 5.20
2 71.20
3 9.00
4 10.11
5 12.21
6 65.00
7 7.00
dtype: float64
Or with axis=1
your_dict = {
'key1': [10, 100.1, 0.98, 1.2],
'key2': [72.5],
'key3': [1, 5.2, 71.2, 9, 10.11, 12.21, 65, 7]
}
pd.concat({k: pd.Series(v) for k, v in your_dict.items()}, axis=1)
key1 key2 key3
0 10.00 72.5 1.00
1 100.10 NaN 5.20
2 0.98 NaN 71.20
3 1.20 NaN 9.00
4 NaN NaN 10.11
5 NaN NaN 12.21
6 NaN NaN 65.00
7 NaN NaN 7.00
I suggest you just create a dict of Series, since your keys do not have the same number of values:
{ key: pd.Series(val) for key, val in x.items() }
You can then do Pandas operations on each column individually.
Once you have that, if you really want a DataFrame, you can:
pd.DataFrame({ key: pd.Series(val) for key, val in x.items() })
key1 key2 key3
0 10.00 72.5 1.00
1 100.10 NaN 5.20
2 0.98 NaN 71.20
3 1.20 NaN 9.00
4 NaN NaN 10.11
5 NaN NaN 12.21
6 NaN NaN 65.00
7 NaN NaN 7.00