Data Analysis with Microsoft Excel: Updated for Office 2007

Chapter 11 Times Series 443

surprising that high temperatures are correlated with the high temperatures

of the previous year. In such a series, if an observation is above the mean,

then its neighboring observations are also likely to be above the mean,

and the autocorrelations with nearby observations are high. In fact, when

there is a trend, the autocorrelations tend to remain high even for high lag

numbers.

STATPLUS TIPS

You can use StatPlus’s ACF(range, lag) function to compute

autocorrelations for specifi c lag values. Here, range is the range

of cells containing the time series data, and lag is the number

of observations to lag. Note that values must be placed within a

single column.

Other ACF Patterns

Other time series show different types of autocorrelation patterns. Figure 11-8

shows four examples of time series (trend, cyclical, oscillating, and random),

along with their associated autocorrelation functions.

•

Figure 11-8

Four sample

time series with
corresponding
ACF patterns

You have already seen the fi rst example with the temperature data. The

trend need not be increasing; a decreasing trend also produces the type of

ACF pattern shown in the fi rst example in Figure 11-8.

The seasonal or cyclical pattern shown in the second example is common

in weather data that follows a seasonal pattern (such as monthly average