In linear regression analysis, the errors are
assumed to be random in nature. But in case of time series, it is
often violated by the autocorrelation property present within
the data. This is rectified by SARIMAX, which explains the
autocorrelation (Cools et al., 2009).