Statsmodels: Statistical Modeling and Econometrics
Overview
Statsmodels is Python's premier library for statistical modeling, providing tools for estimation, inference, and diagnostics across a wide range of statistical methods. Apply this skill for rigorous statistical analysis, from simple linear regression to complex time series models and econometric analyses.
When to Use This Skill
This skill should be used when:
- Fitting regression models (OLS, WLS, GLS, quantile regression)
- Performing generalized linear modeling (logistic, Poisson, Gamma, etc.)
- Analyzing discrete outcomes (binary, multinomial, count, ordinal)
- Conducting time series analysis (ARIMA, SARIMAX, VAR, forecasting)
- Running statistical tests and diagnostics
- Testing model assumptions (heteroskedasticity, autocorrelation, normality)
- Detecting outliers and influential observations
- Comparing models (AIC/BIC, likelihood ratio tests)
- Estimating causal effects
- Producing publication-ready statistical tables and inference
Quick Start Guide
Linear Regression (OLS)
import statsmodels.api as sm
import numpy as np
import pandas as pd
# Prepare data - ALWAYS add constant for intercept
X = sm.add_constant(X_data)
# Fit OLS model
model = sm.OLS(y, X)
results = model.fit()
# View comprehensive results
print(results.summary())
# Key results
print(f"R-squared: {results.rsquared:.4f}")
print(f"Coefficients:\\n{results.params}")
print(f"P-values:\\n{results.pvalues}")
# Predictions with confidence intervals
predictions = results.get_prediction(X_new)
pred_summary = predictions.summary_frame()
print(pred_summary) # includes mean, CI, prediction intervals
# Diagnostics
from statsmodels.stats.diagnostic import het_breuschpagan
bp_test = het_breuschpagan(results.resid, X)
print(f"Breusch-Pagan p-value: {bp_test[1]:.4f}")
# Visualize residuals
import matplotlib.pyplot as plt
plt.scatter(results.fittedvalues, results.resid)
plt.axhline(y=0, color='r', linestyle='--')
plt.xlabel('Fitted values')
plt.ylabel('Residuals')
plt.show()
Logistic Regression (Binary Outcomes)
from statsmodels.discrete.discrete_model import Logit
# Add constant
X = sm.add_constant(X_data)
# Fit logit model
model = Logit(y_binary, X)
results = model.fit()
print(results.summary())
# Odds ratios
odds_ratios = np.exp(results.params)
print("Odds ratios:\\n", odds_ratios)
# Predicted probabilities
probs = results.predict(X)
# Binary predictions (0.5 threshold)
predictions = (probs > 0.5).astype(int)
# Model evaluation
from sklearn.metrics import classification_report, roc_auc_score
print(classification_report(y_binary, predictions))
print(f"AUC: {roc_auc_score(y_binary, probs):.4f}")
# Marginal effects
marginal = results.get_margeff()
print(marginal.summary())
Time Series (ARIMA)
from statsmodels.tsa.arima.model import ARIMA
from statsmodels.graphics.tsaplots import plot_acf, plot_pacf
# Check stationarity
from statsmodels.tsa.stattools import adfuller
adf_result = adfuller(y_series)
print(f"ADF p-value: {adf_result[1]:.4f}")
if adf_result[1] > 0.05:
# Series is non-stationary, difference it
y_diff = y_series.diff().dropna()
# Plot ACF/PACF to identify p, q
fig, (ax1, ax2) = plt.subplots(2, 1, figsize=(12, 8))
plot_acf(y_diff, lags=40, ax=ax1)
plot_pacf(y_diff, lags=40, ax=ax2)
plt.show()
# Fit ARIMA(p,d,q)
model = ARIMA(y_series, order=(1, 1, 1))
results = model.fit()
print(results.summary())
# Forecast
forecast = results.forecast(steps=10)
forecast_obj = results.get_forecast(steps=10)
forecast_df = forecast_obj.summary_frame()
print(forecast_df) # includes mean and confidence intervals
# Residual diagnostics
results.plot_diagnostics(figsize=(12, 8))
plt.show()
Generalized Linear Models (GLM)
import statsmodels.api as sm
# Poisson regression for count data
X = sm.add_constant(X_data)
model = sm.GLM(y_counts, X, family=sm.families.Poisson())
results = model.fit()
print(results.summary())
# Rate ratios (for Poisson with log link)
rate_ratios = np.exp(results.params)
print("Rate ratios:\\n", rate_ratios)
# Check overdispersion
overdispersion = results.pearson_chi2 / results.df_resid
print(f"Overdispersion: {overdispersion:.2f}")
if overdispersion > 1.5:
# Use Negative Binomial instead
from statsmodels.discrete.count_model import NegativeBinomial
nb_model = NegativeBinomial(y_counts, X)
nb_results = nb_model.fit()
print(nb_results.summary())
Core Statistical Modeling Capabilities
1. Linear Regression Models
Comprehensive suite of linear models for continuous outcomes with various error structures.
Available models:
- OLS: Standard linear regression with i.i.d. errors
- WLS: Weighted least squares for heteroskedastic errors
- GLS: Generalized least squares for arbitrary covariance structure
- GLSAR: GLS with autoregressive errors for time series
- Quantile Regression: Conditional quantiles (robust to outliers)
- Mixed Effects: Hierarchical/multilevel models with random effects
- Recursive/Rolling: Time-varying parameter estimation
Key features:
- Comprehensive diagnostic tests
- Robust standard errors (HC, HAC, cluster-robust)
- Influence statistics (Cook's distance, leverage, DFFITS)
- Hypothesis testing (F-tests, Wald tests)
- Model comparison (AIC, BIC, likelihood ratio tests)
- Prediction with confidence and prediction intervals
When to use: Continuous outcome variable, want inference on coefficients, need diagnostics
Reference: See references/linear_models.md for detailed guidance on model selection, diagnostics, and best practices.
2. Generalized Linear Models (GLM)
Flexible framework extending linear models to non-normal distributions.
Distribution families:
- Binomial: Binary outcomes or proportions (logistic regression)
- Poisson: Count data
- Negative Binomial: Overdispersed counts
- Gamma: Positive continuous, right-skewed data
- Inverse Gaussian: Positive continuous with specific variance structure
- Gaussian: Equivalent to OLS
- Tweedie: Flexible family for semi-continuous data
Link functions:
- Logit, Probit, Log, Identity, Inverse, Sqrt, CLogLog, Power
- Choose based on interpretation needs and model fit
Key features:
- Maximum likelihood estimation via IRLS
- Deviance and Pearson residuals
- Goodness-of-fit statistics
- Pseudo R-squared measures
- Robust standard errors
When to use: Non-normal outcomes, need flexible variance and link specifications
Reference: See references/glm.md for family selection, link functions, interpretation, and diagnostics.
3. Discrete Choice Models
Models for categorical and count outcomes.
Binary models:
- Logit: Logistic regression (odds ratios)
- Probit: Probit regression (normal distribution)
Multinomial models:
- MNLogit: Unordered categories (3+ levels)
- Conditional Logit: Choice models with alternative-specific variables
- Ordered Model: Ordinal outcomes (ordered categories)
Count models:
- Poisson: Standard count model
- Negative Binomial: Overdispersed counts
- Zero-Inflated: Excess zeros (ZIP, ZINB)
- Hurdle Models: Two-stage models for zero-heavy data
Key features:
- Maximum likelihood estimation
- Marginal effects at means or average marginal effects
- Model comparison via AIC/BIC
- Predicted probabilities and classification
- Goodness-of-fit tests
When to use: Binary, categorical, or count outcomes
Reference: See references/discrete_choice.md for model selection, interpretation, and evaluation.
4. Time Series Analysis
Comprehensive time series modeling and forecasting capabilities.
Univariate models:
- AutoReg (AR): Autoregressive models
- ARIMA: Autoregressive integrated moving average
- SARIMAX: Seasonal ARIMA with exogenous variables
- Exponential Smoothing: Simple, Holt, Holt-Winters
- ETS: Innovations state space models
Multivariate models:
- VAR: Vector autoregression
- VARMAX: VAR with MA and exogenous variables
- Dynamic Factor Models: Extract common factors
- VECM: Vector error correction models (cointegration)
Advanced models:
- State Space: Kalman filtering, custom specifications
- Regime Switching: Markov switching models
- ARDL: Autoregressive distributed lag
Key features:
- ACF/PACF analysis for model identification
- Stationarity tests (ADF, KPSS)
- Forecasting with prediction intervals
- Residual diagnostics (Ljung-Box, heteroskedasticity)
- Granger causality testing
- Impulse response functions (IRF)
- Forecast error variance decomposition (FEVD)
When to use: Time-ordered data, forecasting, understanding temporal dynamics
Reference: See references/time_series.md for model selection, diagnostics, and forecasting methods.
5. Statistical Tests and Diagnostics
Extensive testing and diagnostic capabilities for model validation.
Residual diagnostics:
- Autocorrelation tests (Ljung-Box, Durbin-Watson, Breusch-Godfrey)
- Heteroskedasticity tests (Breusch-Pagan, White, ARCH)
- Normality tests (Jarque-Bera, Omnibus, Anderson-Darling, Lilliefors)
- Specification tests (RESET, Harvey-Collier)
Influence and outliers:
- Leverage (hat values)
- Cook's distance
- DFFITS and DFBETAs
- Studentized residuals
- Influence plots
Hypothesis testing:
- t-tests (one-sample, two-sample, paired)
- Proportion tests
- Chi-square tests
- Non-parametric tests (Mann-Whitney, Wilcoxon, Kruskal-Wallis)
- ANOVA (one-way, two-way, repeated measures)
Multiple comparisons:
- Tukey's HSD
- Bonferroni correction
- False Discovery Rate (FDR)
Effect sizes and power:
- Cohen's d, eta-squared
- Power analysis for t-tests, proportions
- Sample size calculations
Robust inference:
- Heteroskedasticity-consistent SEs (HC0-HC3)
- HAC standard errors (Newey-West)
- Cluster-robust standard errors
When to use: Validating assumptions, detecting problems, ensuring robust inference
Reference: See references/stats_diagnostics.md for comprehensive testing and diagnostic procedures.
Formula API (R-style)
Statsmodels supports R-style formulas for intuitive model specification:
import statsmodels.formula.api as smf
# OLS with formula
results = smf.ols('y ~ x1 + x2 + x1:x2', data=df).fit()
# Categorical variables (automatic dummy coding)
results = smf.ols('y ~ x1 + C(category)', data=df).fit()
# Interactions
results = smf.ols('y ~ x1 * x2', data=df).fit() # x1 + x2 + x1:x2
# Polynomial terms
results = smf.ols('y ~ x + I(x**2)', data=df).fit()
# Logit
results = smf.logit('y ~ x1 + x2 + C(group)', data=df).fit()
# Poisson
results = smf.poisson('count ~ x1 + x2', data=df).fit()
# ARIMA (not available via formula, use regular API)
Model Selection and Comparison
Information Criteria
# Compare models using AIC/BIC
models = {
'Model 1': model1_results,
'Model 2': model2_results,
'Model 3': model3_results
}
comparison = pd.DataFrame({
'AIC': {name: res.aic for name, res in models.items()},
'BIC': {name: res.bic for name, res in models.items()},
'Log-Likelihood': {name: res.llf for name, res in models.items()}
})
print(comparison.sort_values('AIC'))
# Lower AIC/BIC indicates better model
Likelihood Ratio Test (Nested Models)
# For nested models (one is subset of the other)
from scipy import stats
lr_stat = 2 * (full_model.llf - reduced_model.llf)
df = full_model.df_model - reduced_model.df_model
p_value = 1 - stats.chi2.cdf(lr_stat, df)
print(f"LR statistic: {lr_stat:.4f}")
print(f"p-value: {p_value:.4f}")
if p_value < 0.05:
print("Full model significantly better")
else:
print("Reduced model preferred (parsimony)")
Cross-Validation
from sklearn.model_selection import KFold
from sklearn.metrics import mean_squared_error
kf = KFold(n_splits=5, shuffle=True, random_state=42)
cv_scores = []
for train_idx, val_idx in kf.split(X):
X_train, X_val = X.iloc[train_idx], X.iloc[val_idx]
y_train, y_val = y.iloc[train_idx], y.iloc[val_idx]
# Fit model
model = sm.OLS(y_train, X_train).fit()
# Predict
y_pred = model.predict(X_val)
# Score
rmse = np.sqrt(mean_squared_error(y_val, y_pred))
cv_scores.append(rmse)
print(f"CV RMSE: {np.mean(cv_scores):.4f} ± {np.std(cv_scores):.4f}")
Best Practices
Data Preparation
- Always add constant: Use
sm.add_constant() unless excluding intercept
- Check for missing values: Handle or impute before fitting
- Scale if needed: Improves convergence, interpretation (but not required for tree models)
- Encode categoricals: Use formula API or manual dummy coding
Model Building
- Start simple: Begin with basic model, add complexity as needed
- Check assumptions: Test residuals, heteroskedasticity, autocorrelation
- Use appropriate model: Match model to outcome type (binary→Logit, count→Poisson)
- Consider alternatives: If assumptions violated, use robust methods or different model
Inference
- Report effect sizes: Not just p-values
- Use robust SEs: When heteroskedasticity or clustering present
- Multiple comparisons: Correct when testing many hypotheses
- Confidence intervals: Always report alongside point estimates
Model Evaluation
- Check residuals: Plot residuals vs fitted, Q-Q plot
- Influence diagnostics: Identify and investigate influential observations
- Out-of-sample validation: Test on holdout set or cross-validate
- Compare models: Use AIC/BIC for non-nested, LR test for nested
Reporting
- Comprehensive summary: Use
.summary() for detailed output
- Document decisions: Note transformations, excluded observations
- Interpret carefully: Account for link functions (e.g., exp(β) for log link)
- Visualize: Plot predictions, confidence intervals, diagnostics
Common Workflows
Workflow 1: Linear Regression Analysis
- Explore data (plots, descriptives)
- Fit initial OLS model
- Check residual diagnostics
- Test for heteroskedasticity, autocorrelation
- Check for multicollinearity (VIF)
- Identify influential observations
- Refit with robust SEs if needed
- Interpret coefficients and inference
- Validate on holdout or via CV
Workflow 2: Binary Classification
- Fit logistic regression (Logit)
- Check for convergence issues
- Interpret odds ratios
- Calculate marginal effects
- Evaluate classification performance (AUC, confusion matrix)
- Check for influential observations
- Compare with alternative models (Probit)
- Validate predictions on test set
Workflow 3: Count Data Analysis
- Fit Poisson regression
- Check for overdispersion
- If overdispersed, fit Negative Binomial
- Check for excess zeros (consider ZIP/ZINB)
- Interpret rate ratios
- Assess goodness of fit
- Compare models via AIC
- Validate predictions
Workflow 4: Time Series Forecasting
- Plot series, check for trend/seasonality
- Test for stationarity (ADF, KPSS)
- Difference if non-stationary
- Identify p, q from ACF/PACF
- Fit ARIMA or SARIMAX
- Check residual diagnostics (Ljung-Box)
- Generate forecasts with confidence intervals
- Evaluate forecast accuracy on test set
Reference Documentation
This skill includes comprehensive reference files for detailed guidance:
references/linear_models.md
Detailed coverage of linear regression models including:
- OLS, WLS, GLS, GLSAR, Quantile Regression
- Mixed effects models
- Recursive and rolling regression
- Comprehensive diagnostics (heteroskedasticity, autocorrelation, multicollinearity)
- Influence statistics and outlier detection
- Robust standard errors (HC, HAC, cluster)
- Hypothesis testing and model comparison
references/glm.md
Complete guide to generalized linear models:
- All distribution families (Binomial, Poisson, Gamma, etc.)
- Link functions and when to use each
- Model fitting and interpretation
- Pseudo R-squared and goodness of fit
- Diagnostics and residual analysis
- Applications (logistic, Poisson, Gamma regression)
references/discrete_choice.md
Comprehensive guide to discrete outcome models:
- Binary models (Logit, Probit)
- Multinomial models (MNLogit, Conditional Logit)
- Count models (Poisson, Negative Binomial, Zero-Inflated, Hurdle)
- Ordinal models
- Marginal effects and interpretation
- Model diagnostics and comparison
references/time_series.md
In-depth time series analysis guidance:
- Univariate models (AR, ARIMA, SARIMAX, Exponential Smoothing)
- Multivariate models (VAR, VARMAX, Dynamic Factor)
- State space models
- Stationarity testing and diagnostics
- Forecasting methods and evaluation
- Granger causality, IRF, FEVD
references/stats_diagnostics.md
Comprehensive statistical testing and diagnostics:
- Residual diagnostics (autocorrelation, heteroskedasticity, normality)
- Influence and outlier detection
- Hypothesis tests (parametric and non-parametric)
- ANOVA and post-hoc tests
- Multiple comparisons correction
- Robust covariance matrices
- Power analysis and effect sizes
When to reference:
- Need detailed parameter explanations
- Choosing between similar models
- Troubleshooting convergence or diagnostic issues
- Understanding specific test statistics
- Looking for code examples for advanced features
Search patterns:
# Find information about specific models
grep -r "Quantile Regression" references/
# Find diagnostic tests
grep -r "Breusch-Pagan" references/stats_diagnostics.md
# Find time series guidance
grep -r "SARIMAX" references/time_series.md
Common Pitfalls to Avoid
- Forgetting constant term: Always use
sm.add_constant() unless no intercept desired
- Ignoring assumptions: Check residuals, heteroskedasticity, autocorrelation
- Wrong model for outcome type: Binary→Logit/Probit, Count→Poisson/NB, not OLS
- Not checking convergence: Look for optimization warnings
- Misinterpreting coefficients: Remember link functions (log, logit, etc.)
- Using Poisson with overdispersion: Check dispersion, use Negative Binomial if needed
- Not using robust SEs: When heteroskedasticity or clustering present
- Overfitting: Too many parameters relative to sample size
- Data leakage: Fitting on test data or using future information
- Not validating predictions: Always check out-of-sample performance
- Comparing non-nested models: Use AIC/BIC, not LR test
- Ignoring influential observations: Check Cook's distance and leverage
- Multiple testing: Correct p-values when testing many hypotheses
- Not differencing time series: Fit ARIMA on non-stationary data
- Confusing prediction vs confidence intervals: Prediction intervals are wider
Getting Help
For detailed documentation and examples:
1---2author: luo-kai3name: statsmodels-24description: Statistical models library for Python. Use when you need specific model classes (OLS, GLM, mixed models, ARIMA) with detailed diagnostics, residuals, and inference. Best for econometrics, time series, rigorous inference with coefficient tables. For guided statistical test selection with APA reporting use statistical-analysis.5license: BSD-3-Clause license6metadata:7 source: https://skillmd.com/skills/luokai0/statsmodels-28 skill-author: K-Dense Inc.9---1011# Statsmodels: Statistical Modeling and Econometrics1213## Overview1415Statsmodels is Python's premier library for statistical modeling, providing tools for estimation, inference, and diagnostics across a wide range of statistical methods. Apply this skill for rigorous statistical analysis, from simple linear regression to complex time series models and econometric analyses.1617## When to Use This Skill1819This skill should be used when:20- Fitting regression models (OLS, WLS, GLS, quantile regression)21- Performing generalized linear modeling (logistic, Poisson, Gamma, etc.)22- Analyzing discrete outcomes (binary, multinomial, count, ordinal)23- Conducting time series analysis (ARIMA, SARIMAX, VAR, forecasting)24- Running statistical tests and diagnostics25- Testing model assumptions (heteroskedasticity, autocorrelation, normality)26- Detecting outliers and influential observations27- Comparing models (AIC/BIC, likelihood ratio tests)28- Estimating causal effects29- Producing publication-ready statistical tables and inference3031## Quick Start Guide3233### Linear Regression (OLS)3435```python36import statsmodels.api as sm37import numpy as np38import pandas as pd3940# Prepare data - ALWAYS add constant for intercept41X = sm.add_constant(X_data)4243# Fit OLS model44model = sm.OLS(y, X)45results = model.fit()4647# View comprehensive results48print(results.summary())4950# Key results51print(f"R-squared: {results.rsquared:.4f}")52print(f"Coefficients:\\n{results.params}")53print(f"P-values:\\n{results.pvalues}")5455# Predictions with confidence intervals56predictions = results.get_prediction(X_new)57pred_summary = predictions.summary_frame()58print(pred_summary) # includes mean, CI, prediction intervals5960# Diagnostics61from statsmodels.stats.diagnostic import het_breuschpagan62bp_test = het_breuschpagan(results.resid, X)63print(f"Breusch-Pagan p-value: {bp_test[1]:.4f}")6465# Visualize residuals66import matplotlib.pyplot as plt67plt.scatter(results.fittedvalues, results.resid)68plt.axhline(y=0, color='r', linestyle='--')69plt.xlabel('Fitted values')70plt.ylabel('Residuals')71plt.show()72```7374### Logistic Regression (Binary Outcomes)7576```python77from statsmodels.discrete.discrete_model import Logit7879# Add constant80X = sm.add_constant(X_data)8182# Fit logit model83model = Logit(y_binary, X)84results = model.fit()8586print(results.summary())8788# Odds ratios89odds_ratios = np.exp(results.params)90print("Odds ratios:\\n", odds_ratios)9192# Predicted probabilities93probs = results.predict(X)9495# Binary predictions (0.5 threshold)96predictions = (probs > 0.5).astype(int)9798# Model evaluation99from sklearn.metrics import classification_report, roc_auc_score100101print(classification_report(y_binary, predictions))102print(f"AUC: {roc_auc_score(y_binary, probs):.4f}")103104# Marginal effects105marginal = results.get_margeff()106print(marginal.summary())107```108109### Time Series (ARIMA)110111```python112from statsmodels.tsa.arima.model import ARIMA113from statsmodels.graphics.tsaplots import plot_acf, plot_pacf114115# Check stationarity116from statsmodels.tsa.stattools import adfuller117118adf_result = adfuller(y_series)119print(f"ADF p-value: {adf_result[1]:.4f}")120121if adf_result[1] > 0.05:122 # Series is non-stationary, difference it123 y_diff = y_series.diff().dropna()124125# Plot ACF/PACF to identify p, q126fig, (ax1, ax2) = plt.subplots(2, 1, figsize=(12, 8))127plot_acf(y_diff, lags=40, ax=ax1)128plot_pacf(y_diff, lags=40, ax=ax2)129plt.show()130131# Fit ARIMA(p,d,q)132model = ARIMA(y_series, order=(1, 1, 1))133results = model.fit()134135print(results.summary())136137# Forecast138forecast = results.forecast(steps=10)139forecast_obj = results.get_forecast(steps=10)140forecast_df = forecast_obj.summary_frame()141142print(forecast_df) # includes mean and confidence intervals143144# Residual diagnostics145results.plot_diagnostics(figsize=(12, 8))146plt.show()147```148149### Generalized Linear Models (GLM)150151```python152import statsmodels.api as sm153154# Poisson regression for count data155X = sm.add_constant(X_data)156model = sm.GLM(y_counts, X, family=sm.families.Poisson())157results = model.fit()158159print(results.summary())160161# Rate ratios (for Poisson with log link)162rate_ratios = np.exp(results.params)163print("Rate ratios:\\n", rate_ratios)164165# Check overdispersion166overdispersion = results.pearson_chi2 / results.df_resid167print(f"Overdispersion: {overdispersion:.2f}")168169if overdispersion > 1.5:170 # Use Negative Binomial instead171 from statsmodels.discrete.count_model import NegativeBinomial172 nb_model = NegativeBinomial(y_counts, X)173 nb_results = nb_model.fit()174 print(nb_results.summary())175```176177## Core Statistical Modeling Capabilities178179### 1. Linear Regression Models180181Comprehensive suite of linear models for continuous outcomes with various error structures.182183**Available models:**184- **OLS**: Standard linear regression with i.i.d. errors185- **WLS**: Weighted least squares for heteroskedastic errors186- **GLS**: Generalized least squares for arbitrary covariance structure187- **GLSAR**: GLS with autoregressive errors for time series188- **Quantile Regression**: Conditional quantiles (robust to outliers)189- **Mixed Effects**: Hierarchical/multilevel models with random effects190- **Recursive/Rolling**: Time-varying parameter estimation191192**Key features:**193- Comprehensive diagnostic tests194- Robust standard errors (HC, HAC, cluster-robust)195- Influence statistics (Cook's distance, leverage, DFFITS)196- Hypothesis testing (F-tests, Wald tests)197- Model comparison (AIC, BIC, likelihood ratio tests)198- Prediction with confidence and prediction intervals199200**When to use:** Continuous outcome variable, want inference on coefficients, need diagnostics201202**Reference:** See `references/linear_models.md` for detailed guidance on model selection, diagnostics, and best practices.203204### 2. Generalized Linear Models (GLM)205206Flexible framework extending linear models to non-normal distributions.207208**Distribution families:**209- **Binomial**: Binary outcomes or proportions (logistic regression)210- **Poisson**: Count data211- **Negative Binomial**: Overdispersed counts212- **Gamma**: Positive continuous, right-skewed data213- **Inverse Gaussian**: Positive continuous with specific variance structure214- **Gaussian**: Equivalent to OLS215- **Tweedie**: Flexible family for semi-continuous data216217**Link functions:**218- Logit, Probit, Log, Identity, Inverse, Sqrt, CLogLog, Power219- Choose based on interpretation needs and model fit220221**Key features:**222- Maximum likelihood estimation via IRLS223- Deviance and Pearson residuals224- Goodness-of-fit statistics225- Pseudo R-squared measures226- Robust standard errors227228**When to use:** Non-normal outcomes, need flexible variance and link specifications229230**Reference:** See `references/glm.md` for family selection, link functions, interpretation, and diagnostics.231232### 3. Discrete Choice Models233234Models for categorical and count outcomes.235236**Binary models:**237- **Logit**: Logistic regression (odds ratios)238- **Probit**: Probit regression (normal distribution)239240**Multinomial models:**241- **MNLogit**: Unordered categories (3+ levels)242- **Conditional Logit**: Choice models with alternative-specific variables243- **Ordered Model**: Ordinal outcomes (ordered categories)244245**Count models:**246- **Poisson**: Standard count model247- **Negative Binomial**: Overdispersed counts248- **Zero-Inflated**: Excess zeros (ZIP, ZINB)249- **Hurdle Models**: Two-stage models for zero-heavy data250251**Key features:**252- Maximum likelihood estimation253- Marginal effects at means or average marginal effects254- Model comparison via AIC/BIC255- Predicted probabilities and classification256- Goodness-of-fit tests257258**When to use:** Binary, categorical, or count outcomes259260**Reference:** See `references/discrete_choice.md` for model selection, interpretation, and evaluation.261262### 4. Time Series Analysis263264Comprehensive time series modeling and forecasting capabilities.265266**Univariate models:**267- **AutoReg (AR)**: Autoregressive models268- **ARIMA**: Autoregressive integrated moving average269- **SARIMAX**: Seasonal ARIMA with exogenous variables270- **Exponential Smoothing**: Simple, Holt, Holt-Winters271- **ETS**: Innovations state space models272273**Multivariate models:**274- **VAR**: Vector autoregression275- **VARMAX**: VAR with MA and exogenous variables276- **Dynamic Factor Models**: Extract common factors277- **VECM**: Vector error correction models (cointegration)278279**Advanced models:**280- **State Space**: Kalman filtering, custom specifications281- **Regime Switching**: Markov switching models282- **ARDL**: Autoregressive distributed lag283284**Key features:**285- ACF/PACF analysis for model identification286- Stationarity tests (ADF, KPSS)287- Forecasting with prediction intervals288- Residual diagnostics (Ljung-Box, heteroskedasticity)289- Granger causality testing290- Impulse response functions (IRF)291- Forecast error variance decomposition (FEVD)292293**When to use:** Time-ordered data, forecasting, understanding temporal dynamics294295**Reference:** See `references/time_series.md` for model selection, diagnostics, and forecasting methods.296297### 5. Statistical Tests and Diagnostics298299Extensive testing and diagnostic capabilities for model validation.300301**Residual diagnostics:**302- Autocorrelation tests (Ljung-Box, Durbin-Watson, Breusch-Godfrey)303- Heteroskedasticity tests (Breusch-Pagan, White, ARCH)304- Normality tests (Jarque-Bera, Omnibus, Anderson-Darling, Lilliefors)305- Specification tests (RESET, Harvey-Collier)306307**Influence and outliers:**308- Leverage (hat values)309- Cook's distance310- DFFITS and DFBETAs311- Studentized residuals312- Influence plots313314**Hypothesis testing:**315- t-tests (one-sample, two-sample, paired)316- Proportion tests317- Chi-square tests318- Non-parametric tests (Mann-Whitney, Wilcoxon, Kruskal-Wallis)319- ANOVA (one-way, two-way, repeated measures)320321**Multiple comparisons:**322- Tukey's HSD323- Bonferroni correction324- False Discovery Rate (FDR)325326**Effect sizes and power:**327- Cohen's d, eta-squared328- Power analysis for t-tests, proportions329- Sample size calculations330331**Robust inference:**332- Heteroskedasticity-consistent SEs (HC0-HC3)333- HAC standard errors (Newey-West)334- Cluster-robust standard errors335336**When to use:** Validating assumptions, detecting problems, ensuring robust inference337338**Reference:** See `references/stats_diagnostics.md` for comprehensive testing and diagnostic procedures.339340## Formula API (R-style)341342Statsmodels supports R-style formulas for intuitive model specification:343344```python345import statsmodels.formula.api as smf346347# OLS with formula348results = smf.ols('y ~ x1 + x2 + x1:x2', data=df).fit()349350# Categorical variables (automatic dummy coding)351results = smf.ols('y ~ x1 + C(category)', data=df).fit()352353# Interactions354results = smf.ols('y ~ x1 * x2', data=df).fit() # x1 + x2 + x1:x2355356# Polynomial terms357results = smf.ols('y ~ x + I(x**2)', data=df).fit()358359# Logit360results = smf.logit('y ~ x1 + x2 + C(group)', data=df).fit()361362# Poisson363results = smf.poisson('count ~ x1 + x2', data=df).fit()364365# ARIMA (not available via formula, use regular API)366```367368## Model Selection and Comparison369370### Information Criteria371372```python373# Compare models using AIC/BIC374models = {375 'Model 1': model1_results,376 'Model 2': model2_results,377 'Model 3': model3_results378}379380comparison = pd.DataFrame({381 'AIC': {name: res.aic for name, res in models.items()},382 'BIC': {name: res.bic for name, res in models.items()},383 'Log-Likelihood': {name: res.llf for name, res in models.items()}384})385386print(comparison.sort_values('AIC'))387# Lower AIC/BIC indicates better model388```389390### Likelihood Ratio Test (Nested Models)391392```python393# For nested models (one is subset of the other)394from scipy import stats395396lr_stat = 2 * (full_model.llf - reduced_model.llf)397df = full_model.df_model - reduced_model.df_model398p_value = 1 - stats.chi2.cdf(lr_stat, df)399400print(f"LR statistic: {lr_stat:.4f}")401print(f"p-value: {p_value:.4f}")402403if p_value < 0.05:404 print("Full model significantly better")405else:406 print("Reduced model preferred (parsimony)")407```408409### Cross-Validation410411```python412from sklearn.model_selection import KFold413from sklearn.metrics import mean_squared_error414415kf = KFold(n_splits=5, shuffle=True, random_state=42)416cv_scores = []417418for train_idx, val_idx in kf.split(X):419 X_train, X_val = X.iloc[train_idx], X.iloc[val_idx]420 y_train, y_val = y.iloc[train_idx], y.iloc[val_idx]421422 # Fit model423 model = sm.OLS(y_train, X_train).fit()424425 # Predict426 y_pred = model.predict(X_val)427428 # Score429 rmse = np.sqrt(mean_squared_error(y_val, y_pred))430 cv_scores.append(rmse)431432print(f"CV RMSE: {np.mean(cv_scores):.4f} ± {np.std(cv_scores):.4f}")433```434435## Best Practices436437### Data Preparation4384391. **Always add constant**: Use `sm.add_constant()` unless excluding intercept4402. **Check for missing values**: Handle or impute before fitting4413. **Scale if needed**: Improves convergence, interpretation (but not required for tree models)4424. **Encode categoricals**: Use formula API or manual dummy coding443444### Model Building4454461. **Start simple**: Begin with basic model, add complexity as needed4472. **Check assumptions**: Test residuals, heteroskedasticity, autocorrelation4483. **Use appropriate model**: Match model to outcome type (binary→Logit, count→Poisson)4494. **Consider alternatives**: If assumptions violated, use robust methods or different model450451### Inference4524531. **Report effect sizes**: Not just p-values4542. **Use robust SEs**: When heteroskedasticity or clustering present4553. **Multiple comparisons**: Correct when testing many hypotheses4564. **Confidence intervals**: Always report alongside point estimates457458### Model Evaluation4594601. **Check residuals**: Plot residuals vs fitted, Q-Q plot4612. **Influence diagnostics**: Identify and investigate influential observations4623. **Out-of-sample validation**: Test on holdout set or cross-validate4634. **Compare models**: Use AIC/BIC for non-nested, LR test for nested464465### Reporting4664671. **Comprehensive summary**: Use `.summary()` for detailed output4682. **Document decisions**: Note transformations, excluded observations4693. **Interpret carefully**: Account for link functions (e.g., exp(β) for log link)4704. **Visualize**: Plot predictions, confidence intervals, diagnostics471472## Common Workflows473474### Workflow 1: Linear Regression Analysis4754761. Explore data (plots, descriptives)4772. Fit initial OLS model4783. Check residual diagnostics4794. Test for heteroskedasticity, autocorrelation4805. Check for multicollinearity (VIF)4816. Identify influential observations4827. Refit with robust SEs if needed4838. Interpret coefficients and inference4849. Validate on holdout or via CV485486### Workflow 2: Binary Classification4874881. Fit logistic regression (Logit)4892. Check for convergence issues4903. Interpret odds ratios4914. Calculate marginal effects4925. Evaluate classification performance (AUC, confusion matrix)4936. Check for influential observations4947. Compare with alternative models (Probit)4958. Validate predictions on test set496497### Workflow 3: Count Data Analysis4984991. Fit Poisson regression5002. Check for overdispersion5013. If overdispersed, fit Negative Binomial5024. Check for excess zeros (consider ZIP/ZINB)5035. Interpret rate ratios5046. Assess goodness of fit5057. Compare models via AIC5068. Validate predictions507508### Workflow 4: Time Series Forecasting5095101. Plot series, check for trend/seasonality5112. Test for stationarity (ADF, KPSS)5123. Difference if non-stationary5134. Identify p, q from ACF/PACF5145. Fit ARIMA or SARIMAX5156. Check residual diagnostics (Ljung-Box)5167. Generate forecasts with confidence intervals5178. Evaluate forecast accuracy on test set518519## Reference Documentation520521This skill includes comprehensive reference files for detailed guidance:522523### references/linear_models.md524Detailed coverage of linear regression models including:525- OLS, WLS, GLS, GLSAR, Quantile Regression526- Mixed effects models527- Recursive and rolling regression528- Comprehensive diagnostics (heteroskedasticity, autocorrelation, multicollinearity)529- Influence statistics and outlier detection530- Robust standard errors (HC, HAC, cluster)531- Hypothesis testing and model comparison532533### references/glm.md534Complete guide to generalized linear models:535- All distribution families (Binomial, Poisson, Gamma, etc.)536- Link functions and when to use each537- Model fitting and interpretation538- Pseudo R-squared and goodness of fit539- Diagnostics and residual analysis540- Applications (logistic, Poisson, Gamma regression)541542### references/discrete_choice.md543Comprehensive guide to discrete outcome models:544- Binary models (Logit, Probit)545- Multinomial models (MNLogit, Conditional Logit)546- Count models (Poisson, Negative Binomial, Zero-Inflated, Hurdle)547- Ordinal models548- Marginal effects and interpretation549- Model diagnostics and comparison550551### references/time_series.md552In-depth time series analysis guidance:553- Univariate models (AR, ARIMA, SARIMAX, Exponential Smoothing)554- Multivariate models (VAR, VARMAX, Dynamic Factor)555- State space models556- Stationarity testing and diagnostics557- Forecasting methods and evaluation558- Granger causality, IRF, FEVD559560### references/stats_diagnostics.md561Comprehensive statistical testing and diagnostics:562- Residual diagnostics (autocorrelation, heteroskedasticity, normality)563- Influence and outlier detection564- Hypothesis tests (parametric and non-parametric)565- ANOVA and post-hoc tests566- Multiple comparisons correction567- Robust covariance matrices568- Power analysis and effect sizes569570**When to reference:**571- Need detailed parameter explanations572- Choosing between similar models573- Troubleshooting convergence or diagnostic issues574- Understanding specific test statistics575- Looking for code examples for advanced features576577**Search patterns:**578```bash579# Find information about specific models580grep -r "Quantile Regression" references/581582# Find diagnostic tests583grep -r "Breusch-Pagan" references/stats_diagnostics.md584585# Find time series guidance586grep -r "SARIMAX" references/time_series.md587```588589## Common Pitfalls to Avoid5905911. **Forgetting constant term**: Always use `sm.add_constant()` unless no intercept desired5922. **Ignoring assumptions**: Check residuals, heteroskedasticity, autocorrelation5933. **Wrong model for outcome type**: Binary→Logit/Probit, Count→Poisson/NB, not OLS5944. **Not checking convergence**: Look for optimization warnings5955. **Misinterpreting coefficients**: Remember link functions (log, logit, etc.)5966. **Using Poisson with overdispersion**: Check dispersion, use Negative Binomial if needed5977. **Not using robust SEs**: When heteroskedasticity or clustering present5988. **Overfitting**: Too many parameters relative to sample size5999. **Data leakage**: Fitting on test data or using future information60010. **Not validating predictions**: Always check out-of-sample performance60111. **Comparing non-nested models**: Use AIC/BIC, not LR test60212. **Ignoring influential observations**: Check Cook's distance and leverage60313. **Multiple testing**: Correct p-values when testing many hypotheses60414. **Not differencing time series**: Fit ARIMA on non-stationary data60515. **Confusing prediction vs confidence intervals**: Prediction intervals are wider606607## Getting Help608609For detailed documentation and examples:610- Official docs: https://www.statsmodels.org/stable/611- User guide: https://www.statsmodels.org/stable/user-guide.html612- Examples: https://www.statsmodels.org/stable/examples/index.html613- API reference: https://www.statsmodels.org/stable/api.html614