CODEWITHSUNDEEP
pythonarrayspandasnumpyrank
Benni
Benni

Reputation: 785

Rank an array: Exclude NaN and assign lowest rank to highest number

I have an array/pandas row:

array = [0.8, np.nan, 0.1, -0.5, 0.7]

I want this output:

array = [1, np.nan, 3, 4, 2]

These methods are ranking in the wrong direction for me:

scipy.stats.mstats.rankdata
scipy.stats.rankdata

Upvotes: 3

Views: 2564

Answers (2)

user2285236
user2285236

Reputation:

Since you mentioned Pandas, you can use Series.rank method:

arr = [0.8, np.nan, 0.1, -0.5, 0.7]
pd.Series(arr).rank(ascending=False)
Out: 
0    1.0
1    NaN
2    3.0
3    4.0
4    2.0
dtype: float64

This creates and returns a Pandas Series. If you want to avoid creating a Series, as @ajcr noted in the comments, you can use the rank function. This returns an ndarray:

pd.algos.rank_1d_float64(arr, ascending=False)
Out: array([  1.,  nan,   3.,   4.,   2.])

Upvotes: 3

Divakar
Divakar

Reputation: 221574

Here's an approach -

mask = ~np.isnan(a)
out = np.full(a.size,np.nan)
out[mask] = np.unique(-a[mask],return_inverse=1)[1]+1

Sample run -

In [48]: a
Out[48]: array([ 0.8,  nan,  0.1, -0.5,  nan,  0.7])

In [49]: out
Out[49]: array([  1.,  nan,   3.,   4.,  nan,   2.])

Upvotes: 1

Related Questions