I'm trying to solve this more complicated question. Here's a smaller problem:
Given df
a b
1 2
5 0
5 9
3 6
1 8
How can I create a column C that is a random selection between the two elements of df['a'] and df['b'] of the same row?
So, given this dummy df, the random operator would choose from the pair (1, 2) for row #1, from (5, 0) for row #2...etc.
Thanks
import random
n = 2 # target row number
random.sample(df.iloc[n, :2], 1) # Pick one number from this row.
For the whole dataframe:
>>> df.loc[:, ['a', 'b']].apply(random.sample, args=(1,), axis=1)
0 [1]
1 [5]
2 [9]
3 [3]
4 [8]
dtype: object
Cleaning it up to extract the single values:
>>> pd.Series([i[0] for i in df.loc[:, ['a', 'b']].apply(random.sample, args=(1,), axis=1)], index=df.index)
0 1
1 5
2 9
3 3
4 8
dtype: int64
Or taking advantage that column 'a' is indexed at zero (False) and column 'b' is indexed at 1 (True):
>>> [df.iat[i, j] for i, j in enumerate(1 * (np.random.rand(len(df)) < .5))]
[1, 5, 5, 6, 8]
No need to use a separate random module:
s = """a b
1 2
5 0
5 9
3 6
1 8
"""
df = pd.read_table(StringIO(s),sep='\s+',engine='python')
df.apply(lambda x: x.sample(n=1).iloc[0],axis=1)
#output:
0 1
1 5
2 9
3 6
4 1
dtype: int64
