Machine Learning With Go

Auto-regressive model assumptions and pitfalls

The main assumptions of auto-regressive models are as follows:

• Stationarity: AR models assume that your time series is stationary. We should not see any trends in the data if we plan on using AR models.
• Ergodicity: This fancy term basically means that the statistical properties of the time series, like mean and variance, should not vary or drift over time.

Whatever time series that we are modeling with AR methods should meet these assumptions. However, even when some data (like our air passenger data) that does not meet these assumptions, we can play some differencing tricks to still take advantage of AR models.