如何用Statsmodels实现含两个随机效应的混合模型？附R语言lmer示例

Got it! To replicate your R mixed-effects model with two random intercepts in Statsmodels, you'll use the MixedLM (Mixed Linear Models) class. Here's how to adapt your existing Python code to match the R model lmer(frequency ~ attitude + (1|subject) + (1|scenario), data=df):

Full Working Code #

import pandas as pd
import statsmodels.regression.mixed_linear_model as smlm

# Load and clean data (your existing code)
df = pd.read_csv('http://www.bodowinter.com/tutorial/politeness_data.csv')
df = df.drop(38)

# Fit the mixed-effects model with two random intercepts
model = smlm.MixedLM.from_formula(
    formula='frequency ~ attitude',
    data=df,
    groups=df['subject'],  # Maps to R's (1|subject) random intercept
    vc_formula={'scenario': '1'}  # Maps to R's (1|scenario) random intercept
)
result = model.fit()

# Print the full results summary
print(result.summary())

Key Details Matching Your R Model #

• The fixed effect structure (frequency ~ attitude) is identical to your R formula.
• groups=df['subject'] defines the first random intercept term (1|subject) as the primary grouping variable.
• vc_formula={'scenario': '1'} adds the second random intercept term (1|scenario) as an additional variance component, which is how Statsmodels handles multiple independent random effects.

What to Expect in the Results #

When you run this code, the output will mirror your R model's output closely:

• Random effects: Variance and standard deviation values for subject, scenario, and the residual (you'll see these labeled under "Variance Components").
• Fixed effects: The intercept estimate and the coefficient for attitudepol, along with their standard errors and t-values, will align with the results you got in R.

内容的提问来源于stack exchange，提问作者HappyPy