00451 ë79a8a0a64f5639f3dddf1b4e40ba63

456

Russell

However, the extract suggests that the same process is underlying the observed magnitude changes in the longer morÄ™ elliptical looking extract. This is because on closer examination, the longer, morÄ™ elliptical extract is morÄ™ like a stretched version of the same "accordion fold." The analyst should notÄ™ though that, for sonie unknown reason, there is a change in the di stance between successively plotted points in the univariate phase map movie. This might signal a shift in the parameter of the underlying process rather than a change in the underlying model. NotÄ™ too that the extract suggests that there is very little noise in this data set.

The third classical data set that will be examined is the Box, Hunter, Hunter (1978) plant yield data set. This data set appears in Table 2.2 of "Statistics for Experimenters" and was being distributed on diskette as a sample data set for a time senes modeling package.

A traditional time order plot of the plant yield data is presented in FigurÄ™ 17 below. In addition, a plot of the autocorrelations is presented in the left half of FigurÄ™ 18 and a plot of the partial autocorrelations is presented in the right half of FigurÄ™ 18.

Upon initial examination of the time order plot in FigurÄ™ 17 below, nothing unusual is noticed.

Upon examination of the plots of the autocorrelations and the partial autocorrelations, there is again nothing unusual to be seen about the data, see FigurÄ™ 18. These three plots form the basie plots that the analyst for this series would examine prior to attempting to fit models to the data.

FigurÄ™ 16. Extracts from the univariate phase map movie of the star magnitude data. The extract on the left is around observation 300. The extract on the right is around observation 240. NotÄ™ that the vertical axes are inverted.