Time Series

CFA Level II
#1

dont have my notes here… what is the numerator when testing for seasonality? denom = 1/sqrt T and how do you draw your conclusion? btw - this will be on the exam 100%…not a wildcard. Time series is referred to in 3 study sessions.

#2

sorry i don’t understand this question

#3

I think you’re talking about how to test for autocorrelations of residuals std error, which is the denominator, is as you said = 1/sqrt T you basically use t-test (statistic is autocorrelation over std err), and see if its significantly different from zero, if it is, then there is seasonality with that lag for example, if autocorrelation for lag 4 has a t-statistic thats significantly different from zero, then you have yearly seasonality (assuming quarterly results)

#4

Also note that you do the seasonality test on the LAST lag’s autocorrelation

#5

Care to explain that further deep2002?

#6

I made a note of it in my secret sauce, I don’t remember where I found it. If I see it, I will post.

#7

I’m guessing you meant on the previous regression output. I’m not sure I’m describing it correctly, but I don’t want to to think it is the last lag on the list.

#8

deep2002 Wrote: ------------------------------------------------------- > Also note that you do the seasonality test on the > LAST lag’s autocorrelation —seasonality can be present on any lag, there is no specific lag you have to pay attention to, you just calculate the t-statistic, if its significant, theres seasonality…thats the way I understand it

#9

Correct seasonality can be on any lag and not just the last one as they have always shown in the books. The so called the December-Sales effect.

#10

Thanks guys for updating me on this.