Here are some links to data sets and data tapers.

• nine time series used extensively as examples in the book
  • individual time series
    1. wind speed (Figure 2)
    2. Willamette River (Figures 2 & 505)
    3. atomic clock (Figure 3)
    4. ocean noise (Figure 3)
    5. simulated AR(2) series (Figure 45)
    6. simulated AR(4) series (Figure 45)
    7. ocean wave (Figure 295)
    8. rough ice profile (Figure 304)
    9. smooth ice profile (Figure 304)
  • all nine time series (these are in one big file on StatLib)
• three other time series used as examples in the book
  1. St. Paul temperatures (Figure 9)
  2. Golden Gate Bridge traffic (Figure 195)
  3. 20 points from rotation of earth series. These numbers are listed on page 287 and are used in Figure 288 to illustrate the computational pathways depicted in Figure 285. Figure 288 has six rows of plots. Each row has two plots. The left-hand plot is the time (or lag) domain representation of a certain quantity, while the right-hand plot is the corresponding frequency domain representation. Each representation consists of 64 numbers, which you can download (along with some comments) via the following six items (one for each row in the figure).
     1. centered time series (padded with 44 zeros) and its DFT
     2. Slepian (dpss) data taper with NW=4 (padded with 44 zeros) and its DFT
     3. tapered time series (padded with 44 zeros) and its DFT
     4. acvs of tapered series (circularly wrapped) and its DFT (unnormalized direct spectral estimate)
     5. Parzen lag window (circularly wrapped) and its DFT
     6. lag windowed acvs of tapered series (circularly wrapped) and its DFT (unnormalized lag window spectral estimate)
     Second example of calculation of a lag window estimate for the 20 points from the rotation of earth series, similar to the one displayed in Figure 288, but now using a Hanning data taper (rather than a Slepian data taper) and a Daniell lag window with m=5 (rather than a Parzen lag window):
     1. centered time series (padded with 44 zeros) and its DFT
     2. Hanning data taper (padded with 44 zeros) and its DFT
     3. tapered time series (padded with 44 zeros) and its DFT
     4. acvs of tapered series (circularly wrapped) and its DFT (unnormalized direct spectral estimate)
     5. Daniell lag window (circularly wrapped) and its DFT
     6. lag windowed acvs of tapered series (circularly wrapped) and its DFT (unnormalized lag window spectral estimate)
• computation of zeroth order dpss data tapers via Equation (386)
  1. modified Bessel function I_0(x), Abramowitz and Stegun (1964), p. 378:
     • constants for Equation 9.8.1
     • constants for Equation 9.8.2
  2. N=16, NW=1 test case
  3. N=20, NW=4 test case (can be used to help reproduce Figure 288)
  4. N=64, NW=3 test case
  5. N=128, NW=4 test case
• dpss multitaper test cases
  1. Figure 106: k=0, k=1, k=2 and k=3 tapers for N=32 and NW=4
  2. Figure 108: k=0, k=1, k=2 and k=3 tapers for N=99 and NW=4
• another simulated AR(4) series created from a simulated white noise sequence consisting of 1024 computer-generated pseudo-random deviates from a Gaussian distribution with zero mean and unit variance (the coefficients for the AR(4) process are as listed on page 46, namely, 2.7607, -3.8106, 2.6535 and -0.9238). These two series can be used for testing a computer implementation of a recipe for simulating AR process available in html, LaTeX, PostScript and PDF formats.
• numerical example that can be used to test the implementation of the Yule-Walker method for estimating the parameters of an autoregressive process
• numerical example that can be used to test the implementation of Burg's algorithm for estimating the parameters of an autoregressive process
• time series used in forthcoming second volume of SAPA
  • bivariate ocean wave series

Also, here is a description of -- and access to -- the sapaclisp archive, a collection of Common Lisp functions that implement most of the procedures in the book (an older version of this archive is on StatLib).