This package simulates and estimates a class of panel regression models of the form, $y_{it} = x_{it}' \beta + w_{it}' \gamma + \epsilon_{it} + \nu_{it}$. The parameter of interest is $\beta$, $x_{it}$ is the vector of endogenous regressors, $w_{it}$ is the vector of exogenous covariates, $\epsilon_{it}$ is the endogenous unobservable, and $\nu_{it}$ is the exogenous unobservable (i.e. measurement error).

The purpose of this package is to provide a convenient estimator of $\beta$ given the most common time-series process assumptions on $\epsilon_{it}$, which are:

1. Fixed Effects (“FE”): $\epsilon_{it} = \bar\epsilon_i, \forall t$.
2. Serially Independent (“iid”): $\epsilon_{it} = \eta_{it}$.
3. Moving Average of Order 1 (“MA1”): $\epsilon_{it} = \eta_{it} + \theta \eta_{it-1}$.
4. First Order Autoregressive (“AR1”): $\epsilon_{it} = \rho \epsilon_{it-1} + \eta_{it}$.

There are only 2 functions in this package:

1. PanelReg(): This function estimates the panel regression model. Documentation is available here or by running ?PanelReg in R.
2. PanelRegSim(): This function simulates data from a panel regression model. Documentation is available here or by running ?PanelRegSim in R.

Before estimation, set up a variable list with the names of your variables:

varnames = list(
  id_name = "ID",
  time_name = "Year",
  outcome_name = "Y",
  endogenous_names = c("X1", "X2"),
  covariate_names = c("W1"),
  fixedeffect_names = c("Group")
)

To estimate the panel regression, the syntax is as follows:

reg = PanelReg(panel_data, panel_model = "MA1", varnames)

The argument panel_model must be one of "exogenous", "FE", "iid", "MA1", or "AR1".

For complete examples and details on the estimators and simulators, see the Get Started page.

To install the package from Github:

devtools::install_github("setzler/PanelReg")

To use the package after it is installed:

library(PanelReg)

Make sure these packages have been installed as well:

library(data.table)
library(fixest)