This series of articles explore the data preparation aspect of time series analysis. Data preparation is often overlooked by analysts, but we believe it is a vital phase that wields a vast influence on the overall analysis and modeling process. The vast majority of time series and econometric theories assume input time series to be stationary and homogenous, with equally-spaced observations and values that are present and real. In practice, we often handle samples with missing values, unequally-spaced observations possible outliers, mean/variance dependency, restricted values ranges and other phenomena. The aim of this series of articles is to address each of these problems and introduce practical methods to overcome them.
Have more questions? Submit a request