#24 Course 4

[Oscar]

Has anyone suggested that this is defective? Or can someone explain why C is the answer?

thanks.

On a personal note of frustration: I got 19/23 and guessed on 17. I do not remember my guesses. However, if I had chosen a "guess letter" I would have gotten 6 more. I had only 1 "D" prior to guessing. I was so panicky; I did not think about that strategy during the exam. I just guessed randomly. Now a retake is probable!

[FoxtrotFool]

The concensus is that C is false because for ARIMA(1,1,1)the residuals are not normal until you take difference once.

Yes, someone has suggested that the question is defective, because on choice E, it says "several residuals rk..." where rk's are residual autocorralations. So there is a confusion of terminations.

[Weatherman]

On 2001-11-12 11:35, FoxtrotFool wrote:
The concensus is that C is false because for ARIMA(1,1,1)the residuals are not normal until you take difference once.

Isn't differencing once part of the ARIMA(1,1,1) specification?

Btw, choice B says initial guesses for the parameters can be determined from the Yule Walker equations. Does anyone know how to do this for the general ARIMA(p,d,q) model?

We have 2 weeks after the exam to appeal for a wrong or defective question.

[Jen]

2 weeks? It's ridiculous. Right is right, wrong is wrong. There should not be a time limit for an appeal. Don't we always say time tells??

[Oscar]

Weatherman,

Exactly! I was anticipating the first response about "differencing may be required...for e's to be white noise.." but for that matter the Yule Walker equations aren't much good then either. Perhaps I should do more investigation into whether the SOA has addressed this.

[Joe]

Stop whining. The worse they can do is cancel it, for people like me who answered C.
Do you see us complaining about cancelling a question that we got right?

OK, the question wasn't 100% clear, but this is what these exams are all about. You should know better.

Shouldn't be an issue if you answered correctly to enough questions.

[Oscar]

Joe,

You make so much sense.
I'm borderline.
THe question is not unclear, (the one with rural and urban drivers was unclear); it's incorrect. Or maybe you can tell me why C is a good answr.

[FoxtrotFool]

Oops, I goofed. From another post...

Residuals e's "resembles" white noise, not constitute white noise.

I think that's just being picky... since strictly speaking, no empirical data can constitute a white noise process.

I also think you are right in that ARIMA isn't the reason, since differencing is part of the construct.

[VP]

Joe

Get a clue. I don't have an answer on #24, but to say it doesn't matter because you should have gotten enough of the other questions is idiotic.

Oscar, if you got 19/23 and guessed on the rest (20% chance on 17) your chances of passing are 69% with a passmark of 22. Probably won't give you peace of mind, but there's no cause for despair either.