Logistic regression in Python

Logit models a probability; the coefficients are log odds.

import numpy as np
import statsmodels.api as sm

rng = np.random.default_rng(8)
practice = rng.uniform(0, 60, size=300)
logits = -3 + 0.09 * practice
passed = rng.binomial(1, 1 / (1 + np.exp(-logits)))

X = sm.add_constant(practice)
model = sm.Logit(passed, X).fit(disp=False)

print(model.params.round(4))
print("odds ratio per session:", round(float(np.exp(model.params[1])), 4))
print(round(model.prsquared, 4), round(model.llf, 2))
print(model.pred_table().astype(int))

How it works

1. np.exp(params) turns a coefficient into an odds ratio.
2. predict returns probabilities, not classes.
3. pred_table is the confusion matrix at a 0.5 threshold.

Keywords and builtins used here

• as
• float
• int
• print
• round

The run, in numbers

Lines

15

Characters to type

473

Tokens

159

Three-star pace

110 tpm

At the three-star pace of 110 tokens a minute, this run takes about 87 seconds.

Type this snippet

Step 1 of 4 in Beyond least squares, step 9 of 19 in Statistics with statsmodels.

← Previous Next →

Logistic regression in other languages

• R