ABSTRACT

Many of the data sets considered in earlier chapters have simply consisted of a

number of values – perhaps concrete strengths based on testing a number of

cylinders, or the number of floods in a particular year, or even the number of deaths

per year due to kicks by horses over a 20-year period. In each case there was no

need to consider the order in which the data were collected. Each data point

provided a single value which could be used to determine a mean or develop a

probability distribution. However, there are data where the time order is important.

Take for example population size: in this situation the size is important, but equally

important is the date at which the estimate was made. Other examples include share

indices from world markets and climatic variations such as increasing sea levels or

carbon dioxide concentration. The changes forecast from these time series, and the

uncertainties inherent in these forecasts, present great challenges for engineers.

Time series are analysed for three reasons: to make forecasts that facilitate

decisions; to simulate realistic scenarios; and to supplement or quantify physical

explanations for phenomena.