14.11.2014 Views

Estimating the Codifference Function of Linear Time Series Models ...

Estimating the Codifference Function of Linear Time Series Models ...

Estimating the Codifference Function of Linear Time Series Models ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

and for p ≠ q<br />

cov(Re(φ 3 (s i , p)), Re(φ 3 (s j , q))) = 2e −σα (|s i| α +|s j| α +|s i−s j| α )<br />

+ 2e −σα (|s i| α +|s j| α +|s i+s j | α) − 4e −σα (2|s i| α +|2s j| α )<br />

cov(Im(φ 1 (s i , p)), Im(φ 1 (s j , q))) = cov(sin(−s i X t ), sin(−s j X t )) = 1 |s i−s j| α<br />

2 {e−σα − e −σα |s i+s j | α }<br />

cov(Im(φ 1 (s i , p)), Im(φ 2 (s j , q))) = cov(Im(φ 2 (s i , p)), Im(φ 1 (s j , q)))<br />

= −cov(Im(φ 1 (s i , p)), Im(φ 1 (s j , q)))<br />

cov(Im(φ 2 (s i , p)), Im(φ 2 (s j , q))) = cov(Im(φ 1 (s i , p)), Im(φ 1 (s j , q)))<br />

cov(Im(φ 3 (s i , p)), Im(φ 3 (s j , q)))<br />

= cov(sin(s i (X t+p − X t )), sin(s j (X t+q − X t ))) + cov(sin(s i (X t+p+q − X t+q )), sin(s j (X t+q − X t )))<br />

+ cov(sin(s i (X t+p+q − X t+p )), sin(s j (X t+p − X t ))) + c pq<br />

Im<br />

where<br />

yielding for p = q<br />

c pq<br />

Im = ⎧<br />

⎨<br />

⎩<br />

cov(Im(φ 3 (s i , k)), Im(φ 3 (s j , k)))<br />

0 if p = q<br />

cov(sin(s i (X t+q − X t+q−p )), sin(s j (X t+q − X t ))) if q > p<br />

cov(sin(s i (X t+p − X t )), sin(s j (X t+p − X t+p−q ))) if p > q<br />

= 1 2 e−2σα |s i−s j| α − 1 2 e−2σα |s i+s j| α + e −σα (|s i| α +|s j| α +|s i+s j | α) − e −σα (|s i| α +|s j| α +|s i−s j | α )<br />

and cov(Im(φ 3 (s i , k)), Im(φ 3 (s j , k))) = 0 for p ≠ q. The o<strong>the</strong>r elements are all zeros. The elements<br />

<strong>of</strong> L k 2 are as given in (34), where <strong>the</strong> elements <strong>of</strong> d k i , i = 1, . . .r are<br />

As from (45) we obtain<br />

and<br />

<strong>the</strong>n <strong>the</strong> (i,j)-th element <strong>of</strong> m RR<br />

kk<br />

d k i (1, 1) = (Re Φ(0, −s i; k)) −1 = e σα |s i| α<br />

d k i (2, 2) = (Re Φ(s i , 0; k)) −1 = e σα |s i| α<br />

d k i (3, 3) = (Re Φ(s i , −s i ; k)) −1 = e 2σα |s i| α )<br />

m RR<br />

kk = cov(Re ˆτ(s, k), Re ˆτ(s, k)) = (I r ⊗ λ 1 )d k V RR<br />

kk dk (I r ⊗ λ T 1 )<br />

m II<br />

kk = cov(Im ˆτ(s, k), Im ˆτ(s, k)) = (I r ⊗ λ 1 )d k V RR<br />

kk d k (I r ⊗ λ T 1 )<br />

and mII<br />

kk<br />

is obtained from<br />

m RR<br />

kk (i, j) = λ 1 d k i V RR<br />

kk (i, j)d k jλ T 1<br />

and<br />

m II<br />

kk (i, j) = λ 1d k i V kk II (i, j)dk j λT 1<br />

which <strong>the</strong>refore after a simple algebra, we obtain<br />

m RR<br />

kk (i, j) = eσα (|s i| α +|s j| α −|s i−s j| α ) { 1<br />

2 eσα (|s i| α +|s j | α −|s i−s j| α) − 1<br />

+ e σα (|s i| α +|s j| α −|s i+s j| α ) { 1<br />

2 eσα (|s i| α +|s j| α −|s i+s j | α) − 1<br />

}<br />

}<br />

+ 1<br />

{<br />

m II<br />

kk (i, j) = (|s i| α +|s j| α −|s i−s j| α ) 1 (|s i| α +|s j| α −|s i−s j | α) }<br />

eσα 2 eσα − 1<br />

{ }<br />

+ e σα (|s i| α +|s j| α −|s i+s j| α )<br />

1 − 1 (|s i| α +|s j| α −|s i+s j | α )<br />

2 eσα<br />

The required result follows directly from (44).<br />

19

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!