Tests for serial correlation of unknown form in dynamic least squares regression with wavelets

On the predictability of crude oil market: A hybrid multiscale wavelet approach

“…The wavelet method is a relatively new tool for time series analysis, but it has been used extensively in the applied areas of economics and finance. For instance, it is used to examine business cycle components in financial and economic time series (see Yogo, 2008); the relationship between inflation and unemployment in macro-and micro-economic contexts (see Gallegati, Gallegati, Ramsey, & Semmler, 2011), growth-inflation linkages (see Uddin, Muzaffar, Arouri, & Sjö, 2017), money supplyinflation dynamics across frequencies and over time (see Bekiros, Muzaffar, Uddin, & Vidal-Garcia, 2017); efficiency of generalized method of moments estimators (see Michis & Sapatinas, 2007), the impact of financial variables on the market (Abry, Flandrin, Taqqu, & Veitch, 2003;Boubaker & Sghaier, 2015;Fan & Gencay, 2010;Gencay, Selcuk, & Whitcher, 2001;Gençay & Signori, 2015;Li & Gençay, 2017;Tan, Chin, & Galagedera, 2014;Xue, Gençay, & Fagan, 2014). Other recent studies where wavelet transformation has been used in forecasting models include, for instance, Michis (2014), where maximal-overlap discrete wavelet transform (MODWT) is used to decompose economic forecasts and their associated forecast errors into different timescales, enabling an evaluation of forecast accuracy across the cycle according to equal-size timescale components of the time series.…”