Panel Data

Analyze data on a cross-section over multiple time points

Panel data consists of observations on multiple subjects collected repeatedly over time. Examples of panel data include data collected on individuals, households, firms, municipalities, states, or countries over the same time period. Panel data analysis can be performed by fitting panel regression models that account for both cross-section effects and time effects and give more reliable parameter estimates compared to linear regression models.

There are two types of panel data:

  • Balanced (complete) panel comprises all observations for each individual are measured at the same time points. Example: Economic data from countries or states collected yearly for 10 years.
  • Unbalanced (incomplete) panel comprises missing observations for some individuals for certain time points. Example: Financial data from firms or individuals where some firms or individuals are older than others.

Common panel regression models include:

  • Panel data fixed-effect models or least squares with dummy variables (LSDV) models: cross-section specific effects are modeled using dummy variables
  • One-way random-effects models: cross-section specific effects are modeled as random-effects
  • Two-way random-effects models: both cross-section effects and time effects are modeled as random effects
  • Nested (hierarchical) models: nested groupings in cross-section data (for example, states nested in countries) are modeled as random effects

Common estimation methods for panel data regression models include:

For more information on how to fit various panel data regression models, see Statistics and Machine Learning Toolbox™, Financial Toolbox™, and Econometrics Toolbox™ for use with MATLAB®.

See also: Statistics and Machine Learning Toolbox, Econometrics Toolbox, Financial Toolbox, linear model, linear regression, regression and ANOVA, Predictive Modeling

Data Analytics

Use MATLAB for Big Data, Machine Learning and Production Analytics Systems.