Time Series Graphics
Updated: February 9, 2006
Jacob: Why do we graph the time series? Why don’t we start with the sample and partial autocorrelations and specify a model?
Rachel: Some items are easier to see in a graph than from numbers. We can separate a time series into eras, detect seasonality, observe changes in drift, volatility, and mean, and see cycles more easily in graphs than in tables of figures.
Jacob: What types of graphs do we form?
Rachel: We use four types of graphs:
We start with scatter plots of interest rates by month. These shows changes in mean, volatility, and drift; help differentiate eras; and show cycles.
Stochasticity overwhelms minor changes. We use moving average charts to eliminate stochasticity and better observe trends.
High stochasticity obscures weak seasonality. We use bar charts to eliminate the stochasticity and better detect seasonality.
Scatter plots do not show sample autocorrelations. We use correlograms to observe the sample autocorrelations.
Jacob: Can we do a student project showing these graphs for various types and time periods of interest rates?
Rachel: The graphs are a part of any student project; they are a tool, not the content of the student project. The student project should deal with some question, such as "What is a reasonable time series model for these interest rates?"