Comparison Of Density Forecasts Tests: Application In Finance And Macroeconomics
Author: Hanzala Zulfiqar

The keen interest of this study leads to capture uncertainty by using evaluation of density forecast techniques and also comparison of evaluation of density forecast tests. Evaluation of density forecast technique by using Probability Integral Transformation (PIT) and evaluation of point forecast technique by using Root Mean Square Error (RMSE) and Mean Square Error (MSE) use for same purpose to identify a best model. In this study we have evaluated the performance of three different tests of Probability Integral Transformation (PIT) on the basis of their size and power through Monte-Carlo simulation. All these tests are better in size and power for a specific data generating processes compared with Berkowitz (2001) LR test as mentioned in Knüppel (2015), Ghosh and Bera (2015) and Raaij and Raunig (2002) respectively and not compared yet. So, in this study Raw Moment Based test, Smooth test and Regression based test have been compared with each other to check their performance through simulation study. In the end, we have conducted an empirical study on different share price time series data (Bank Al-Habib Limited, K-Electric Limited and Maple Leaf Cement Factory Ltd) and exchange rate to identify a univariate model by capturing volatility by using PIT technique. The results of Monte-Carlo simulation indicate that raw moment based test perform well as compared to others. In case of empirical study results show that we cannot rely on a single evaluation of density forecast test to identify a best univariate model. Supervisor:- Dr. Asad Zaman Co-Supervisor:- Dr. Amna Urooj

Meta Data

Keywords : Calibration, JB test, Probability Integral Transformation, Raw Moment Based test, Regression Based test, Sharpness and Monte-Carlo simulation, Smooth test
Supervisor: Asad Zaman

Related Thesis​