NON-DETERMINISTIC MODELING USING QUANTILE REGRESSION

Abstract

In this thesis, we utilize quantile regression to model the conditional quantile of the dependent variable given independent variables to capture more details about the conditional distribution. In addition, we apply the quantile-on-quantile regression model to estimate the impact of an independent variable's quantiles on the conditional quantiles of the dependent variable to uncover the dependence between the independent and dependent variables. We consider the RavenPack news-based index associated with the coronavirus outbreak (Panic, Media Hype, Fake News, Sentiment, Infodemic, and Media Coverage) and the returns of Bitcoin and gold as real-world applications. Our findings demonstrate that the bearish and bullish on Bitcoin and gold are affected by the daily positive and negative shocks in indices caused by coronavirus news asymmetrically. Sentiment induced by coronavirus plays a major role in driving Bitcoin and gold values than other indices. Bitcoin and gold can act as a hedge against coronavirus-related news.