Last week, I gave a presentation on the basics of time series statistics to an informal group of students in biology, forestry, and fisheries who meet every week to share tips and talk over problems involving statistics and R. A time series is just a sequence of observations recorded at regular intervals, one after another. The basic motivation behind time series statistical techniques is the fact that the observations making up a time series usually violate the assumption of independence that underlies most of traditional statistics. Instead of being uncorrelated, they are related to each other, or autocorrelated: what has happened before affects what will happen in the future. Time series models are intended to incorporate this autocorrelation, making it possible to understand the process underlying a time series, and to forecast its future values.
I got a good response to the talk, so I’ve decided to post my slides here, along with an R script including a series of short examples that dovetail with the presentation. I have used a number of these techniques in my own research, and mostly had to teach myself; this presentation is essentially an overview of everything I wish I’d known before I started reading seriously about time series. I’ll emphasize that this is a very cursory overview of an immense field of statistical research and techniques. There’s only so much ground you can cover in a one-hour seminar. The last slide has a list of the books I’ve been working out of for further reading.