00449 ­6b2c81a17b7f60430d45ea04acefc6

454

Russell

FigurÄ™ 13. This univariate phase map movie extract contains two years of data from the senes. Notice that the seasonality and growth are readily apparent. The seasonality is represented by the strong pattem that repeats every 12 observations and the growth is represented by the movement of the pattem to the lower right. As in all the other movie extract plots the vertical axis is inverted.

signal of either a bad data point, an injection of noise into the observations or process, or an actual change in the behavior of the time series. We see nothing of this naturÄ™ that is unusual or suggestive in the univariate phase map movie.

From the airline passenger data set we see that the univariate phase map movie is capable of dealing with both seasonality and growth. In addition the movie is seen to have some potential for diagnostics at least in "high signal" situations.

The second classical data set that we will examine is a time series of about 600 estimates of the magnitude of a variable star. The traditional time order plot of the series is presented in FigurÄ™ 14 below. In addition, the autocorrelation plot is presented on the left side of FigurÄ™ 15 and the partial autocorrelation plot is presented on the right side of FigurÄ™ 15.

The examination of the univariate phase map movie shows the two typical types of extracts from this data set, they are presented in figurÄ™ 16 below. One is a long elliptical pattem and the other is a shorter "accordion like" pattem. Of particular interest is the shorter "accordion like fold" of the extract on the right side of FigurÄ™ 16. These odd looking extracts are from the region in the time order plot where the amplitudÄ™ is lower and the frequency between the peaks seem to be longer.