Feasible weighted projected principal component analysis for semi-parametric factor models
提出一种可行的加权投影主成分分析方法，用于半参数因子模型，能利用协变量更高效地估计因子，并应用于扩散指数预测，用美国债券市场数据验证了其在预测超额债券收益中的表现。
Summary Various factor estimation procedures have been developed, based on the latent factor model. They often consider general conditions that allow for correlations and heteroscedasticity. However, the conventional principal components method does not efficiently estimate the parameters. It also does not accommodate additional covariates, which explain the unknown factors, even if they are available. In particular, a few aggregated macroeconomic variables can be used as covariates in diffusion index forecasts. To account for these features, I propose the feasible weighted projected principal component (WPPC) analysis, based on semi-parametric factor models, and also establish its asymptotic properties. In addition, I apply the WPPC method to the diffusion index forecasting model. Finally, I investigate the performance of the WPPC estimator in forecasting excess bond returns using US bond market and macroeconomic data.