Chapter 2. Understanding Time Series Data
In the previous chapter, we touched upon a general approach of time series analysis which consists of two main steps:
- Data visualization to check the presence of trend, seasonality, and cyclical patterns
- Adjustment of trend and seasonality to generate stationary series
Generating stationary data is important for enhancing the time series forecasting model. Deduction of the trend, seasonal, and cyclical components would leave us with irregular fluctuations which cannot be modeled by using only the time index as an explanatory variable. Therefore, in order to further improve forecasting, the irregular fluctuations are assumed to be independent and identically distributed (iid) observations and modeled by a linear regression on variables other than the time index.
For example, house prices might exhibit both trend and seasonal (for example, quarterly) variations. However, the residuals left after adjusting trend and seasonality might actually depend on...
