Nonparametric and quantile regression approaches: Energy and commodity market links

Nonparametric regression is of great importance with respect to offering great flexibility in data modelling by not requiring a clear description of functional forms for estimated objects, thus relaxing the parametric assumptions imposed on the data generating process. Considering its significance, this dissertation comprises three essays and mainly aims to carry out nonparametric regression, nonparametric quantile estimation and nonparametric quantile regression approaches in order to analyze energy and commodity markets which can be regarded as one of the driving forces affecting the quality of being competitive and growth for contemporary economies. The first essay has aimed to perform the nonparametric causality-in-quantile framework in order to detect the multivariate causations running from crude oil, technology and stock market composite price indexes towards clean energy price indexes and also obtain the average derivative estimates of the conditional quantile function based on series approximations method. The second essay has dealed with predicting the volatility of Bitcoin returns using squared and original returns as proxies for volatility based on the study by Klemelä (2017) -and also performing the quantile estimation- for different prediction horizons. The third essay has explored the dynamic behaviors between gold and the drivers of gold -namely, crude oil and S&P500 stock prices- using the nonparametric vector autoregression approach in predictions and presenting a comparison with some competitive models for 7-day ahead forecasting. Keywords: Nonparametric regression, causality-in-quantile, average derivative, volatility prediction, energy and commodity markets.