Comparison Of Time Domain And Frequency Domain Analysis In Forecasting Sri Lankan Share Market Returns
Time domain analysis and frequency domain analysis are two approaches for analyzing a financial time series. Time domain analysis is widely used in that respect but the latter is applied less. Literature revealed that both approaches have not been applied much in sri lankan context. Also it revealed a knowledge gap in forecasting sri lankan stock market returns. Therefore this study aimed to compare the forecasting ability of time domain and frequency domain in forecasting sri lankan stock returns. Two time domain models: auto regressive integrated moving average (arima) and auto distributed lag model (adlm) were compared with fourier transformation based analysis. Data collection period was year 2011 to 2014. Monthly total market returns and returns of random sample of four business sectors were analyzed. Model assumptions were tested by residual plots, anderson darling test and durbin watson test. Model assessment was based on mean square error (mse). Results revealed that arima and fourier transformation based analysis are suitable in forecasting but adlm is not. Based on the results, it was concluded that frequency domain approach is better than time domain approach in forecasting sri lankan stock market returns.
Keywords- Time domain, Frequency Domain