Abstract:
In the study of time series, new techniques for smoothing the data are very important and evolving. This study presents a new technique for smoothing time series data which is based on average and in the time domain. It is aimed at creating a new series by redistributing the average of the reduced (halved) series of the variable of interest; it has the potential of reducing large peaks and takes care of outliers. The use of contemporary smoothers in modeling time series data are not without some challenges. They are based on assumptions, difficulty in detecting the actual tool to employ, iteratively demanding and so on. This proposed technique is direct and easy to employ and not based on any form of assumptions and iterations. The strength of this technique was tested by comparing its output solely with that of the linear phase filter using both economic data (lending rate, 1984 – 2012) and crop production data (1990 – 2012) using the Mean Absolute Deviation (MAD), Mean Square Error (MSE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE) as criteria. The test of normality was also performed on the errors using Kolmogorov – Smirnov test. From the results obtained, the residual ɛt from the technique was found to be white noise, which is the building block for time series models. It also showed that the proposed technique yields minimum values for the various error measures and fit a normal curve better than the linear phase filter for both data sets.
