Generalized dynamic principal component for monthly nonstationary stock market price in technology sector

The majority of stock market price is nonstationary, while only few have stationary pattern. It is noted that past researches usually transformed the stock market price into stationary prior to analysis which may lead to the loss of data originality. Thus, a direct application of the nonstationary stock market price is of main interest in this study, as such generalized dynamic principal component (GDPC) performs the analysis directly without transformation. As well as, Brillinger dynamic principal component (BDPC) were also used on the nonstationary stock market price for comparison. This dataset consists of the most recent five-year monthly observations of six different regions in technology sector. Stationarity test was performed prior to the application and the analyses were carried out based on the reconstruction of lags of the time series. The results showed that the GDPC for six stock market prices have lower mean squared error compared to BDPC. Also, the percentage of explained variance in the first component were much higher in GDPC. Thus, this indicated that GDPC model is more suitable for prediction compared to its counterpart.