covariance stationary from regression table

CFA Level II
#1

Hi, from an output of regression results… what factors do we look for to determine if it’s covariance stationary?

Do we look at the regression coefficients/t-stat or the autocorrelation information or the first differenced correlation/coefficient results?

thanks

#2

If it’s an AR(1) model, |b1| must be less than 1 for it to be covariance stationary.

That’s the only model mentioned in the CFA curriculum.

1 Like
#3

Dickey-Fuller test. I don’t believe you can infer covariance stationary just by looking at a generic regression output.

1 Like
#4

The Dickey-Fuller test is a test for a unit root. You would use it if b1 is close to 1. It won’t necessarily tell you whether or not the regression is covariance stationary.

#5

Does the autocorrelation results of the residual for AR(1) infer imply anything? (or is it just for detecting serial correlation?)

thanks

#6

i believe it’s only for detecting serial correlation which would imply the accuracy of the model

#7

Hence if b-sub1 was 1 or greater, the dependent variable would not be mean reverting, correct?

#8

Correct.

Also if it’s less than −1.

1 Like
#9

Read the original question wrong :slight_smile:

#10

I’m shocked. Stunned. Floored. My world, as I (thought that I) knew it has come to a crashing end.

#11

I missed the “not” in “…not mean reverting…” which then led me to misinterpret the answers!