Newton-Raphson Optimization for Generalized Heckman Model Estimation
Arguments
- start
A numeric vector of initial parameter guesses for the selection, outcome, dispersion, and correlation equations.
- YS
A binary vector indicating selection status (1 if selected, 0 otherwise).
- XS
A matrix of independent variables for the selection equation.
- YO
A numeric vector of observed outcomes (dependent variable) for the outcome equation.
- XO
A matrix of independent variables for the outcome equation.
- Msigma
A matrix representing the predictors for the dispersion parameter.
- Mrho
A matrix representing the predictors for the correlation parameter.
- w
A numeric vector of observation weights, used in the likelihood computation.
Value
A list with the following components:
- coefficients
Named vector of estimated coefficients for selection, outcome, dispersion, and correlation equations.
- fitted.values
Named list with fitted values for each equation (selection, outcome, dispersion, correlation).
- residuals
Numeric vector of residuals for the selection and outcome equations.
- loglik
Log-likelihood value of the fitted model.
- vcov
Variance-covariance matrix of the estimated parameters.
- aic
Akaike Information Criterion (AIC) for the model.
- bic
Bayesian Information Criterion (BIC) for the model.
- optimization
Details of the optimization process, including convergence information.
Details
This function estimates the parameters of a generalized Heckman selection model using a Newton-Raphson optimization algorithm. It supports the modeling of selection and outcome equations, along with associated dispersion and correlation structures.
This function uses the Newton-Raphson algorithm to estimate the parameters of
a generalized Heckman model, which accounts for sample selection bias.
The model is composed of a selection equation (modeled by YS and XS), an
outcome equation (modeled by YO and XO), and additional equations for
dispersion (Msigma) and correlation (Mrho). The optimization process
maximizes the log-likelihood of the model, allowing for robust estimation of
selection bias, while also estimating associated dispersion and correlation
parameters.
The function outputs the coefficients, fitted values, residuals, and several information criteria for model comparison.