Python中sm.Logit逻辑回归：提取.summary更多结果及模型统计信息

Got it, let's walk through how to get the standard error, z-statistic, p-values, and that top section of model-level statistics from your sm.Logit results in statsmodels.

First: Variable-Level Statistics (std err, z, p-value) #

Once you've fit your model (let's assume you have a results object from model.fit()), you can directly access these stats via built-in attributes—no need to parse the summary text:

import statsmodels.api as sm
import pandas as pd

# Example model fit (adjust with your X/y data)
model = sm.Logit(y, sm.add_constant(X))
results = model.fit()

• Standard Error: Use results.bse — this returns a pandas Series with variable names mapped to their standard errors, matching the order of results.params.
• Z-Statistic: Use results.tvalues (note: even though it's named tvalues, Logit models use z-statistics here; statsmodels just reuses the attribute name for consistency).
• P-Value: Use results.pvalues — this gives the p-value for each variable's coefficient significance test.

To combine all these into a clean, easy-to-read DataFrame (along with your existing beta values and confidence intervals):

summary_table = pd.DataFrame({
    'Beta': results.params,
    'Std Err': results.bse,
    'Z': results.tvalues,
    'P-Value': results.pvalues,
    'CI Lower': results.conf_int()[0],
    'CI Upper': results.conf_int()[1]
})
print(summary_table)

Second: Model-Level Summary Statistics (the top section of .summary()) #

The first table in the .summary() output has overall model metrics like log-likelihood, pseudo R-squared, AIC, etc. You can access each metric individually, or pull the entire formatted table as text:

Access individual model metrics: #

• Log-Likelihood: results.llf
• Null Log-Likelihood (model with only intercept): results.llnull
• Pseudo R-squared: results.prsquared
• AIC: results.aic
• BIC: results.bic
• Chi-squared statistic (for model vs. null comparison): results.chi2
• P-value for chi-squared test: results.chi2pvalue

Get the full top summary table as text: #

If you want the exact same formatted table as .summary() outputs, use:

top_summary_table = results.summary().tables[0].as_text()
print(top_summary_table)

This will print the entire first section, including headers and all model stats, just like you see in the .summary() output.

内容的提问来源于stack exchange，提问作者Joe