Time Series Concepts
A time series is a sequence of observations taken at evenly spaced time in-
tervals. The sequence could be daily temperature measurements, weekly
sales fi gures, monthly stock market prices, quarterly profi ts, or yearly power-
consumption data. Time series analysis involves looking for patterns that
help us understand what is happening with the data and help us predict
future observations. For some time series data (for example, monthly sales
fi gures), you can identify patterns that change with the seasons. This sea-
sonal behavior is important in forecasting.
Usually the best way to start analyzing a time series is by plotting the data
against time to show trends, seasonal patterns, and outliers. If the variability
of the series changes with time, the series might benefi t from a transformation
that stabilizes the variance. Constant variance is assumed in much of time se-
ries analysis, just as in regression and analysis of variance, so it pays to see
fi rst whether a transformation is needed. The logarithmic transformation is
one such example that is especially useful for economic data. For example,
if there is growth in power consumption over the years, then the month-to-
month variation might also increase proportionally. In this case, it might be
useful to analyze either the log or the percentage change, which should have
a variance that changes little over time.Time Series Example: The Rise in
Global Temperatures
To illustrate these ideas, you’ve been provided the Global Temperature work-
book (Source: http://data.giss.nasa.gov/gistemp/tabledata/GLB.Ts.txt). The
workbook contains average annual temperature readings compiled by NASA,
covering the years 1880 through 1997. The NASA data are often used by
climatologists investigating climate change and global warming. Table 11-1
describes the range names and data contained in the workbook.Table 11-1 Global Temperature Workbook
Range Name Range Description
Year A2:A129 The year
Decade B2:B129 The decade
Celsius C2:C129 The average annual global temperature in
degrees Celsius
Fahrenheit D2:D129 The average annual global temperature in
degrees Fahrenheit432 Statistical Methods