i.i.d exam relevance

[Gary2584]

Hi David,

I just wanted to confirm how testable is the i.i.d concept in the P1 exam, i have gone through the 1.a.1 Intro to VaR_v3 spreadsheet, but when i started the 1.b.5 practice bag i got confused with the first part of the sheet as it talks about i.i.d assumptions.

Please let me know that in VaR (foundations concept) what are the topics that are likely to be tested.

Thanks,
Gary

 

[David Harper CFA FRM]

Hi Gary,

Thank you for noticing, as I did intend to color red "Assumptions: i.i.d., normal returns" to introduce the typical parametric VaR calculations because, while people tend to remember the normal assumption (maybe b/c it's in the media), they seem to forget i.i.d.

i.i.d. is not itself an assumption that tends to be directly quizzed, although it has.

Nevertheless, it is extremely foundational and important. Primarily:

1. A "random sample" (which implicates much of our econometrics) is technically a set of i.i.d. random variables (drawn from a population). It's such a key assumption here that it's hard to overstate (just from yesterday http://forum.bionicturtle.com/threads/p1-t2-212-difference-between-two-means.5357/)
2. In has a popular role in distribution; e.g., a binomial [used to backtest VaR, albeit that is P2] assumes i.i.d. Bernoulli. Poisson assumes i.i.d.
3. Most importantly, and w.r.t to the practice bag, i.i.d. is required to scale volatility/VaR with the square root rule. When we scale a 1-day VaR to 10-day VaR with *SQRT(10), or an annual VaR to a 10-day VaR with *SQRT(10/250), we assume i.i.d. where one of the "i" is "independent" such that nonzero auto- or serial-correlation violates the i.i.d. assumption. Normality is not the precondition, the precondition for the SRR is i.i.d. returns with finite variance.

The FRM P1 won't ask a lot of questions about i.i.d. directly, rather several questions are likely to include it as an assumption (as a practical matter, a topic like VaR almost must be introduced with the i.i.d. returns assumption, before relaxing it with more sophisticated approaches), i hope that helps, thanks,

 

[Gary2584]

Hi David,

Thanks for the reply, it cleared my doubts, one more thing the value of confidence level will be given in the exam or else we can assume 99% =2.33 and 95% = 1.64 and proceed with the calculation.

Thanks,
Gary

 

[David Harper CFA FRM]

Hi Gary, sure np. You cannot assume VaR confidence (nor horizon), however, the exam will generally ask about either 95% or 99% confidence and will expect you to know the corresponding one-tailed, normal deviates of 1.645 (or 1.65) and 2.33.

(P2 needs to know that basel internal market risk calibrates a 10-day 99% and credit/operational use one-year 99.9%, but as the latter are not normal even there does not expect you to know the 99.9% deviate), thanks,

 

[chouchouc]

Hello David, I keep reading contradictory things about iid. For the part 3 of the exam, could you give me a list of Var methodologies that assume iid and that dont. thanks a lot