Title |
Alternative functional restrictions for midas regression / |
Translation of Title |
Alternatyvūs funkciniai apribojimai MIDAS regresijai. |
Authors |
Liktas, Mindaugas |
Full Text |
|
Pages |
33 |
Keywords [eng] |
MIDAS, HAR, HARCH, realized volatility, Gamma restriction function. |
Abstract [eng] |
In this work, MIDAS regression model is used as a main tool to forecast realized volatility. Classical HAR model and normalized exponential Almon polynomial are used as benchmark functional restrictions to compare other MIDAS models. New restriction functions are introduced as alternatives to HAR and Almon polynomial. Alternatives for HAR model were chosen according to HARCH model, where different market components with different time horizons can be expressed as a power of two natural numbers. Some empirical results lead to an inclusion of restriction function as a combination of two Gamma probability density functions. Estimation results show that newly proposed functional constraints quite often show better forecasting performance regarding mean squared errors. There are also indexes for which new restriction functions are worse or don’t fit at all, what means that one must not choose new restriction functions blindly against classical ones all the time. |
Dissertation Institution |
Vilniaus universitetas. |
Type |
Master thesis |
Language |
English |
Publication date |
2017 |