Penalized quadratic inference functions for single-index models with longitudinal data (Q958914)

The authors consider the single index model \[ E(y|{\mathbf{x}})=h_0\big({\mathbf{x}}^{\prime}{\pmb{\beta}}_0\big), \] where \(y\) is a dependent scalar variable, \({\mathbf{x}} \in \mathbb{R}_d\) is a vector of covariates, \({\pmb{\beta}}_0\in \mathbb{R}_{d}\) is a vector of unknown parameters with the restriction \(||{\pmb{\beta}}_0||=1\) and \(h(\cdot)\) is unknown univariate link function. The data are expected to be of a longitudinal type. Section~2 provides an estimating procedure under the single-index model for the longitudinal data. In Section~3 several practical issues the authors faced when suggesting the estimation procedure are discussed. Consistency and asymptotic normality are proven in Section~4. In Section~5, a non-parametric goodness-of-fit test on the unknown link function is obtained. Finally, Monte Carlo simulations and real data analysis assessing the finite sample performance of the proposed procedure are described in Section~6.