Abstract:
|
In the era of big data, researchers interested in developing statistical models are challenged with how to achieve parsimony. In the field of macroeconomic, macro-forecasting requires researchers to make inference on multiple and inter-dependencies time series data. The dependencies on macroeconomic time series data have a significant effect on decision making and forecasting accuracy. I develop an optimal and novel sufficient dimension reduction approach to handle the ultra-high dimensional macroeconomic series data. Specifically, my method aims to select the most significant variables without loss of information in the data.
|