Simulated Gaussian copula endogeneity data

A synthetic dataset generated to illustrate Gaussian copula endogeneity corrections. The dependent variable is driven by an endogenous regressor that is correlated with the structural error via a shared noise component.

Usage

sim_endog

Format

A data frame with 2000 rows and 4 variables:

y

Outcome variable

z_endog

Endogenous regressor that is correlated with the error term

x_exog

Exogenous regressor

w_instr

Instrument used by copula estimators

Details

The true data-generating process is $$y = 1 + 2 z_{\text{endog}} + 1.5 x_{\text{exog}} + u,$$ where a latent Gaussian factor induces correlation between \(z_{\text{endog}}\) and the structural error \(u\). The instrument \(w_{\text{instr}}\) shifts the endogenous regressor but is excluded from the outcome equation.

Examples

data(sim_endog)
str(sim_endog)
#> 'data.frame':	2000 obs. of  4 variables:
#>  $ y      : num  0.0945 4.2063 -2.7204 3.1972 2.1352 ...
#>  $ z_endog: num  0.3616 0.8219 -1.5403 1.3684 -0.0897 ...
#>  $ x_exog : num  -0.699 0.996 -0.693 -0.103 0.604 ...
#>  $ w_instr: num  1.482 1.705 -0.934 0.605 -0.524 ...