How could I run multiple univariate regressions by group with sklearn?

Question

I am trying to replicate this solution Python pandas: how to run multiple univariate regression by group but using sklearn linear regression instead of statsmodels.

import pandas as pd
import numpy as np
from sklearn.linear_model import LinearRegression

df = pd.DataFrame({
  'y': np.random.randn(20),
  'x1': np.random.randn(20), 
  'x2': np.random.randn(20),
  'grp': ['a', 'b'] * 10})


def ols_res(x, y):
    return pd.Series(LinearRegression.fit(x,y).predict(x))


results = df.groupby('grp').apply(lambda x : x[['x1', 'x2']].apply(ols_res, y=x['y']))

print(results)

I get:

TypeError: ("fit() missing 1 required positional argument: 'y'", 'occurred at index x1')

The results should be the same as the article I linked, which is:

             x1        x2
grp                      
a   0 -0.102766 -0.205196
    1 -0.073282 -0.102290
    2  0.023832  0.033228
    3  0.059369 -0.017519
    4  0.003281 -0.077150
        ...       ...
b   5  0.072874 -0.002919
    6  0.180362  0.000502
    7  0.005274  0.050313
    8 -0.065506 -0.005163
    9  0.003419 -0.013829

score 1 · Accepted Answer · answered Apr 09 '19 at 03:02

There are a two minor problems with your code:

You don't instantiate a LinearRegression object, so your code actually tries to call the unbound fit method of the LinearRegression class.
Even if you fix this, the LinearRegression instance will be unable to perform fit and transform because it expects a 2D array and gets a 1D one. Accordingly, you also need to reshape the array contained in each Series.

import pandas as pd
import numpy as np
from sklearn.linear_model import LinearRegression

df = pd.DataFrame({
  'y': np.random.randn(20),
  'x1': np.random.randn(20), 
  'x2': np.random.randn(20),
  'grp': ['a', 'b'] * 10})

def ols_res(x, y):
    x_2d = x.values.reshape(len(x), -1)
    return pd.Series(LinearRegression().fit(x_2d, y).predict(x_2d))

results = df.groupby('grp').apply(lambda df: df[['x1', 'x2']].apply(ols_res, y=df['y']))

print(results)

Output:

             x1        x2
grp                      
a   0 -0.126680  0.137907
    1 -0.441300 -0.595972
    2 -0.285903 -0.385033
    3 -0.252434  0.560938
    4 -0.046632 -0.718514
    5 -0.267396 -0.693155
    6 -0.364425 -0.476643
    7 -0.221493 -0.779082
    8 -0.203781  0.722860
    9 -0.106912 -0.090262
b   0 -0.015384  0.092137
    1  0.478447  0.032881
    2  0.366102  0.059832
    3 -0.055907  0.055388
    4 -0.221876  0.013941
    5 -0.054299  0.048263
    6  0.043979  0.024594
    7 -0.307831  0.059972
    8 -0.226570 -0.024809
    9  0.394460  0.038921

How could I run multiple univariate regressions by group with sklearn?

1 Answers1