A1448
Title: Unobserved heterogeneity in factor-augmented panel quantile model
Authors: Xinbing Kong - Nanjing Audit University (China)
Wei Wang - Shandong University (China) [presenting]
Xiaodong Yan - Shandong University (China)
Abstract: Panel data models with various structural patterns, e.g., group pattern, structural break, sparsity, received increasingly more attention in statistics and econometrics. A factor-augmented panel quantile model is proposed that combines panel quantile model with factor structure that allows high-dimensional distribution-specific factors with loadings with structural breaks and sparsity. In addition, the slopes allow for unobserved grouped structures across individuals. An integrative procedure that detects the information regarding sparsity, group and structural break patterns of factor loadings and variable selection on high-dimensional covariates simultaneously via a doubly penalized hinge loss function. We use a speed iterative coordinate descent algorithm that automatically integrates structural break and group pattern factor loadings pertaining to a common one and recovers the sparsity formation of regression coefficients and loading elements. Consistency and asymptotic normality for the proposed estimators are developed. We show that the resulting estimators exhibit oracle properties, i.e., the proposed estimator is asymptotically equivalent to the oracle estimator obtained using the known sparsity, group and structural break patterns. Furthermore, the simulation studies provide supportive evidence that the proposed method has good finite sample performance A real data empirical application has been provided to highlight the proposed method.
