Correlation and Regression

I came across a question in schweser which requires us to test the significance of regression coefficients with 34 degrees of freedom. However, the table provided in the textbook has no row for 34 degrees of freedom. Should I be prepared to come across such a situation on the exam? If yes, what is the best way to go about it?

Do you mean the critical value table doesn’t have a row for 34?

On the exam, they likely won’t give you something like this (my guess). I suspect they would give your 36 df if the closest table value is 35 (since 35 is a more conservative number due to a larger critical value threshold).

You probably won’t get answers that conflict on the exam such as fail to reject at 30 df (larger critical value) and reject at 35 df (smaller critical value) I’d like to think if they gave you 34 df that you’d make the same decision at each of the provided df critical values so that your answer is clear.

Maybe @S2000 can weigh in on what he has seen.

Yes, I’m talking about the critical value table. The one in the schweser textbook jumps from 30 to 40

Do you make the same conclusion irrespective of whether you use 30 vs 40 df?

If so, not to worry on this question.

In the exam, the CFAI will give you a table if needed or it will be painfully obvious like a t-stat of 25 with 150 df (since you’d have to recall that around 120 df on a t-distribution is about 1.96 at the 5% two tailed critical level). That’s my suspicion. In a pinch on the exam where you have no clue and the jump in critical values gives different conclusions, you could average the two critical values. It’s not exact, but better than nothing.

Edited out something that is very likely beyond the scope of the exam (don’t recall how often they use CI for hypothesis testing alternative).

Yeah, the conclusion is the same, but it just gave rise to the possibility of their being a scenario where the conclusion may not be the same, and I do remember something about interpolation being mentioned at level 1.

Anyway, I guess you’re right, it does seem highly unlikely that this issue would come up on the actual exam.

I’d be very surprised if you had to interpolate a critical value on the exam.

That’s not the purpose of the exam.