In Notes Book1 Page 236 and 237. The example (started on bottom of p235) says it used 40 quarters number. When it use AR(1), the # of observations is 39(within the ANOVA on p236). I understand this one. Then in the ANOVA on p237, it says it include a seasonal lag. Then the # of observations became 38… Why it’s 38 not 36? How the 38 was calculated out?

you lose 1 observation for every lag

still confused it is said AR(2) will lose two observations, while AR(4) should lose four. And for things (on page 237) like: ln y(t)=b0 + b1 ln y(t-1) + b2 ln y(t-4) is it AR(2) or AR(4)?

ln y(t)=b0 + b1 ln y(t-1) + b2 ln y(t-4) that’s an AR(2) with lags at 1 and 4. you lose 2 observations

Thanks, Slash. I still got problem on the definition of AR§ in this q as in the notes, p224 on book1 there is: AR model of order p, AR§, is expressed as: xt = b0 + b1 x(t-1) + b2 x(t-2) + … + bp x(t-p) +e(t) so is it saying tha when p=4 and b2=b3=b(p-1)=0, it became AR model of order 2? According to the definition shouldn’t it still be AR(4)? As I see this one, it loses one observation in AR(1) because you need to get input from 1 to T-1 for y(t-1) instead of 1 to T. then if you include a seasonal lag, shouldn’t you use 1 to T-4 for y(t-4) instead of 1 to T? why it’s T-2?