||This dissertation is comprised of three essays that apply Factor Augmented Vector Autoregression (FAVAR) with large macroeconomic datasets of the U.S., Hong Kong and Chinese economies. The first essay applies the FAVAR to investigate the appropriateness frequent adjustments of the policy interest rate in a prolonged manner. Economic activities implied by impulse response functions from hypothetical scenarios, which assume less frequent monetary policy, are compared with those generated from actual fed policies under the record of Alan Greenspan (1987-2006). The essay finds that a less frequent monetary policy approach could control inflation with less negative impact on real economic activities, and major economic variables would be less volatile in the long term. The second essay applies the FAVAR model to forecast GDP growth rate, unemployment rate and inflation rate of the Hong Kong economy. There is no factor model forecasting literature on the Hong Kong economy. The objective is to find out whether factor forecasting of using a large dataset can improve forecast performance of the Hong Kong economy. To avoid misspecification of the number of factors in the FAVAR, combination forecasts are constructed. It is found that forecasts from FAVAR model overall outperform simple VAR and AR models, especially when forecasting horizon increases. Generally, combination forecasts solve the misspecification problem. The third essay compares the effectiveness of quantity-based and price-based monetary policies in China using FAVAR. This essay is the pioneer to identify the 1-year lending rate and deposit rate as the policy rates, and includes yield curve information in the analysis. It is found that effects of tightening monetary policies in China follow largely the stylized facts of long run neutrality of money on real activities, a long term fall in inflation and a short term rise in interest rates.