Components of Time Series
- Trend (Secular Trend)
- Cyclical variation and Seasonal variation
- Irregular component
Additive and Multiplicative Models
- Additive Model: In this, assumption is time series data Y=T+C+S+I.
- Multiplicative Model: The assumption is time series data Y=T*C*S*I
As nicely explained in this blog, the implicit assumption under additive model is that, for a monthly data if the difference between let us say, January and July is constant over years. However, in multiplicative model, not the absolute difference but the proportion (or percentage change) is assumed to be constant over years. Those who are interested can visit this blog as well.
Time series Decomposition in Python
To demonstrate time series decomposition, I downloaded the airline passengers data from this site. There are two ways of decomposing time series data.
- Using seasonal_decompose method: We can specify additive or multiplicative model for decomposing the time series data. Here the time series is decomposed in to trend-cyclical component, seasonal component and irregular component as shown in the picture below.
- Using UnobservedComponents method: Here we can decompose a time series data into trend, seasonal, cyclical and irregular components.
If you have any questions or suggestions, feel free to share. I will be happy to interact.