# Time Series

I just read that you can use Durbin Watson to test for Serial Corr in a linear trend and log linear trend model…this is not good. So, in an AR model you can’t use DW correct?

no you have to use t test.

1/square root of T

3_letters Wrote: ------------------------------------------------------- > I just read that you can use Durbin Watson to test > for Serial Corr in a linear trend and log linear > trend model…this is not good. > > So, in an AR model you can’t use DW correct? And sure as sh!t they are going to conveniently provide you with the Durbin Watson statistic…

caspian Wrote: ------------------------------------------------------- > 1/square root of T with t being n-(k+1) right?

I thought it was just 1/square root n (or t) and not the degrees of freedom

You’re right. I’m going through a qbank ? right now (12798) where you have to calc a confidence interval and it has me confused.

Niblita cannot use DW to test serial correlation on autoregressive models. the equation you mention is used to determine errors which is used to look for serial correlation. calculating 1sqroot of n is the error term gives you the t-stat and if they are greater than 2 they are statistically different than 0, would would mean the model is not correctly specified (serial correlation exists)

I never said you use DW.

I just spent a precious 2 hrs going thru the Time Series reading in the CFAi books and feel much better…I could easily get 5/6 on the item set in mock one now…I suggest you guys do the same. It will be on the exam. I can even tell you the limitations to using ARMA model…

DW can not be used to test autocorr when the independent var is a lagged value of dep var

What are the (ARMA) limititations 3L?

eriqnoodle and aeolusloo get a combined -10 for reading comprehension…

Haha, I use that same system when I talk to some people. Just yesterday I was telling someone directions and literally said “Now pay attention because I know you are going to forget something, your attention to detail is a -6.”

What ARMA model? OMG, how come I’ve never heard of? can you refer me to the page? tks. 3_letters Wrote: ------------------------------------------------------- > I just spent a precious 2 hrs going thru the Time > Series reading in the CFAi books and feel much > better…I could easily get 5/6 on the item set > in mock one now…I suggest you guys do the > same. It will be on the exam. > > I can even tell you the limitations to using ARMA > model…

arma isn’t in schwesers… is it in LOS? i’m thinking no. i have looked and read it a couple of times in CFA readings. the arma is easy to find. it’s just last few pages of times series in CFA readings.

thanks so much. I found it. is there anything else that’s not in schweser notes I shld be aware of? btw, do you know where i can find tobin’s q? I think I’ve looked thru all the schweser indexes, and it’s not there. I had asked this Q in other thread, no response. thanks again. westbruin Wrote: ------------------------------------------------------- > arma isn’t in schwesers… is it in LOS? i’m > thinking no. i have looked and read it a couple of > times in CFA readings. > > the arma is easy to find. it’s just last few > pages of times series in CFA readings.

Just read the thing on ARMA, basically its an AR model with a moving average error component. Basically, it is very unstable.

Wow. Sorry for throwing ARMA out there guys. In other news…how much do you know about GARCH?..hahahaha

Zero