hard · FRM Part 1 Quantitative Analysis
A risk manager backtests a 99% one-day VaR model over 250 trading days and observes 8 exceptions. She runs Kupiec's unconditional-coverage (POF) likelihood-ratio test, which is distributed chi^2 with 1 degree of freedom (critical values: 3.84 at 95%, 6.63 at 99%). The test statistic comes out to about 7.73.
What is the correct conclusion?
- Fail to reject the model at both confidence levels, because 8 exceptions is within normal sampling variation around the 2.5 expected
- Reject the model at the 95% level but not at the 99% level, since 7.73 exceeds 3.84 but the stricter test should use the higher exception count cautiously
- Reject the model's accuracy at both the 95% and 99% confidence levels, as 7.73 exceeds both 3.84 and 6.63
- Fail to reject at 95%, because Kupiec's test only flags models that produce too FEW exceptions, not too many
Sign up free to see the explanation and track your rank →
More FRM Part 1 Quantitative Analysis practice
- A probability distribution that is asymmetric and has a significantly long tail extending
- A single discrete trial that results in exactly one of two possible outcomes (success or f
- How does the mean of a lognormal distribution compare to the mean of its associated normal
- If an analyst says a return series has 'fat tails,' what does this imply for a risk model
- If the correlation between two assets is -1.0, what does this indicate about their co-move
- In Bayesian inference, what does the term 'Updating' refer to?
- In combinatorics, which coefficient represents the number of ways to select r items from a
- In the context of credit risk, if D is the event of default and F is a model flag, how is