Göm meny

Errata list

PAGE LINE IN BOOK SHOULD BE
32 -2 summation index N-k+1 \tau = N-k+1
33 2 e^{i\omega\tau} e^{-i\omega\tau}
36 2 e^{i\omega\tau} e^{-i\omega\tau}
36 eq (2.66) Summation index k=1 to M k=0 to M-1
48 10 \int \frac{1}{2\pi i}\int
53 2 R_s(\tau) - R_s^N(\tau) ||R_s(\tau) - R_s^N(\tau)||
54 -7 \Phi_s \Phi_w
65 -7 drived derived
98 -10 s \le 1 s \le t-1
176 16 (eq (6.34a)) G_0(e^{-i\omega}) G_0(e^{-i\xi})
180 -3 (eq (6.48)) \Phi_{uy} \Phi_u
193 -12 ... , = ... ... , M = ...
201 14 (eq (7.21)) summation from k =1 summation from k = 0
211 7 u(t+k) u(t+k-1)
215 -7 M^{-1} N M^{-1}
218 6 [log f]'=f'/f [-log f]'=-f'/f
235 -5 Section 7.8 Section7.7
236 12 (Problem 7G2) Summation to k = N Summation to k = n
249 11 u = w u = r
251 12 {w(t)} {r(t)}
260 5 an and
265 18 (in eq (8.68)) \Phi_{ue} \Phi_{eu}
289 15 (below eq (9.45)) \psi_\rho(t+1) \psi_\rho(t+i)
330 4 ... = - q^{-k}y(t) ... = q^{-k}y(t)
330 2 ... = - q^{-k}u(t) ... = q^{-k}u(t)
330 eq (1054.b) ... = - \frac{D(q)}{C(q)F(q)} ... = \frac{D(q)}{C(q)F(q)})
332 -13 to -15 (3 places) (\beta\varphi - \gamma) (\beta(\varphi - \gamma))
347 10 (in eq (10.110b)) lim 1/N \Pi ... lim X 1/N \Pi ...
350 15 (in eq (10.128b)) O_r ... \hat{O}_r ...
357 19-20 E \bar{E}
436 eq (13.56) \arg\min = \arg\min
456 Problem 13G.4 equation for V_N add a term "N \log \lambda_1 \lambda_2"
456 Problem 13E.3 not possible .. asymptotically if G can be correctly modeled
480 2 aspects. aspects,
481 3 different \ell different f_e
497 -7 identiflability identifiability
505 (16.36) [see below] \sum \frac{\sigma_i^2}{(\sigma_i+\delta)^2} \sum \frac{\sigma_i}{\sigma_i+2\delta}
518 7 Givers Gevers
513 -10 eq (16.63) Switch the transpose on r_M^N
513 -13 element of P element of \lambda*P

Correction and clarification

The expression (8.66) for the spectrum of y is not correct. There is another term. Its integral is however zero, so the discussion following (8.66) about limit estimates is not affected. The simplest way to realize this is to write (8.65) as \epsilon(t)=h(t)+e_0(t). Here h(t) and e_0(t) are uncorrelated, so the variance of \epsilon is equal to the variance of h plus \lambda_0. The spectrum of h is the first term of (8.66), so the discussion can go on from there.

Regarding (16.36) there are two errors in the derivation: First on line 4 page 505, the Hessian is W" = V" + 2 \delta I. Second, on line 7, the first factors H(H+\delta I) should be omitted, and in the last factor \delta I should be replaced by 2\delta I (as just noted). I am grateful to Robert Bos at Delft for sorting this out.

Informationsansvarig: Lennart Ljung
Senast uppdaterad: 2024-08-20