ABSTRACT
Whittle (1953) proposed to approximate La( e, tt, CJ2) by (B 2) I Jrr { 2 ( ) ln(A; J.L) } Lc , J-1, CJ = -2 log CJ h >..; B + ? ( ) d)... 1r -rr u-h )...;() (4.9)
and the estimates
Finally define the (uncentered) periodogram
(4.10)
Following Kashyap and Eom (1988) we can also proceed by regressing log/11 (\) on logh(>.1;6i) over)= 1, ... ,n-I, though this approach leads to less efficient than Gaussian estimates. In fact Ray (1993) used this technique to estimate d3 and d12 in the SCLM process
where the E1 are white noise. Ray (1993) used these estimates as a first step in the estimation of the complete model in Eq. (4.14) for monthly IBM revenues.
