*
* Example 20.3 from pp 628-630
*
open data tablef20-2[1].txt
calendar(q) 1950
data(format=prn,org=columns) 1950:1 1983:4 qtr y m1 p
*
set lrgnp = log(y/p)
set ygrow = 100*(lrgnp-lrgnp{1})
*
* The mean is extracted from the data before computing the periodogram or
* spectrum. Figure 20.2 shows the growth rate with mean extracted.
*
diff(center) ygrow / yc
graph(footer="Figure 20.2 Growth of US Real GNP")
# yc
*
* The periodogram and spectral estimates are produced using the SPECTRUM
* procedure. This, by default, does a log scale graph. NOLOGSCALE is used to
* produce a linear scale graph as shown. To get the periodogram, use the option
* PERIODOGRAM from this. This procedure also has a HEADER and FOOTER options to
* label the graph.
*
@spectrum(periodogram,nologscale,footer="Figure 20.3 Sample Periodogram") yc
*
* Similarly, the log of real gnp has its mean extracted.
*
diff(center) lrgnp / gc
graph(footer="Figure 20.4 Quarterly Data on Real GNP")
# gc
*
* The procedure uses the smoothed periodogram method for computing consistent
* estimates of the spectral density. This isn't described in the text. Smoothed
* periodogram estimators take moving averages of the periodogram ordinates. This
* exploits the assumed smoothness of the underlying spectral density.
*
@spectrum(width=13,nologscale,footer="Figure 20.5 Spectrum for Real GNP") gc