Given a dataframe, I want to get the nonzero values of each row and then find the minimum of absolute values. I want to have a user defined function that does this for me. Also, I do not want to use any for loop since the data is big.
My try
np.random.seed(5) data = np.random.randn(16) mask = np.random.permutation(16)[:6] data[mask] = 0 df = pd.DataFrame(data.reshape(4,4)) 0 1 2 3 0 0.441227 -0.330870 2.430771 0.000000 1 0.000000 1.582481 -0.909232 -0.591637 2 0.000000 -0.329870 -1.192765 0.000000 3 0.000000 0.603472 0.000000 -0.700179 def udf(x): if x != 0: x_min = x.abs().min() return x_min df.apply(udf, axis=1)
I get ValueError: The truth value of a Series is ambiguous. Use a.empty, a.bool(), a.item(), a.any() or a.all().
Question How can I solve the above?
The desired answer is the following:
0.330870 0.591637 0.329870 0.603472
Advertisement
Answer
You can use x.ne(0)
as boolean indexing to filter row
res = df.apply(lambda x: x[x.ne(0)].abs().min(), axis=1)
print(res) 0 0.330870 1 0.591637 2 0.329870 3 0.603472 dtype: float64
Or use min(axis=1)
res = df[df.ne(0)].abs().min(axis=1)