SemiParSampleSelObject: Fitted SemiParSampleSel object
Description
A fitted semiparametric sample selection object returned by function SemiParSampleSel and of class "SemiParSampleSel".Value
- fitList of values and diagnostics extracted from the output of the algorithm. For instance,
fit$gradient and fit$S.h return
the gradient vector and overall penalty matrix scaled by its smoothing parameters, for the bivariate probit model. See the
documentation of trust for details on the diagnostics provided. - gam1Univariate fit for equation 1. See the documentation of
mgcv for full details. - gam2Univariate fit for equation 2.
- gam2.1Univariate fit for equation 2, estimated using an adaptation of the Heckman sample selection correction procedure.
- coefficientsThe coefficients of the fitted semiparametric sample selection model. They are given in the following
order: parametric and regression spline (if present)
coefficients for the first equation, parametric and regression spline
coefficients for the second equation, logarithm of standard deviation of the output (in bivariate normal case) or logarithm of
scale parameter (in normal-gamma case) and transformed association coefficient.
- weightsPrior weights used during model fitting.
- spEstimated smoothing parameters of the smooth components for the fitted sample selection model.
- iter.ifNumber of iterations performed in the initial step of the Fisher scoring algorithm.
- iter.spNumber of iterations performed for the smoothing parameter estimation step.
- iter.fiFinal number of iterations performed after smoothing parameter/leap frog estimation.
- start.vStarting values for all model parameters of the semiparametric sample selection algorithm. These are obtained using the
Heckman sample selection
correction approach when starting values are not provided.
- phiEstimated dispersion for the response of the outcome equation. In the normal bivariate case, this corresponds to the variance.
- sigmaEstimated standard deviation for the response of the outcome equation in the case of normal marginal distribution of the outcome.
- shapeEstimated shape parameter for the response of the outcome equation in the case of gamma marginal distribution of the outcome.
- thetaEstimated coefficient linking the two equations. In the normal bivariate case, this corresponds to the correlation coefficient.
- tauEstimated Kendall's tau association coefficient between the two equations.
- nSample size.
- n.selNumber of selected observations.
- X1Design matrix associated with the first linear predictor.
- X2Design matrix associated with the second linear predictor.
- X1.d2Number of columns of the design matrix for equation 1. This is used for internal calculations.
- X2.d2Number of columns of the design matrix for equation 2.
- l.sp1Number of smooth components in equation 1.
- l.sp2Number of smooth components in equation 2.
- HePenalized hessian.
- HeShUnpenalized hessian.
- VbInverse of the penalized hessian. This corresponds to the Bayesian variance-covariance matrix used for `confidence' interval calculations.
- FThis is given by
Vb*HeSh. - BivDType of bivariate copula distribution employed.
- marginsMargins used in the bivariate copula specification.
- t.edfTotal degrees of freedom of the estimated sample selection model. It is calculated as
sum(diag(F)). - bs.mgfitA list of values and diagnostics extracted from
magic in mgcv. - conv.spIf
TRUE then the smoothing parameter selection algorithm converged. - wor.cWorking model quantities given by the square root of
the weight matrix times the pseudo-data vector and design matrix,
rW.Z and rW.X. - eta1,eta2Estimated linear predictors for the two equations.
- y1Binary outcome of the selection equation.
- y2Dependent variable of the outcome equation.
- logLValue of the (unpenalized) log-likelihood evaluated at the (penalized or unpenalized) parameter
estimates.
- fpIf
TRUE, then a fully parametric model was fitted.