Skip to contents

Christoffersen conditional-coverage test for VaR exceedances

Source: R/k05_tranche1.R
morie_christoffersen_cc.Rd

Splits "i.i.d. Bernoulli(alpha)" into its two testable halves and adds them: LR_cc = LR_uc + LR_ind, chi-square on 2 df. LR_ind compares a first-order Markov chain on the indicator against an i.i.d. one, so it fires when breaches CLUSTER even if their overall rate is right – the failure Kupiec's test alone cannot see.

Usage

morie_christoffersen_cc(hits, alpha = 0.05)

Arguments

hits

0/1 exceedance indicator, in time order.

alpha

the VaR tail probability the model claims.

Value

list: statistic (LR_cc), pvalue, lr_uc, pvalue_uc, lr_ind, pvalue_ind, transition counts n00..n11, df, method.

References

Christoffersen, P. F. (1998). Evaluating interval forecasts. International Economic Review 39(4), 841-862. Cross-checked against rugarch's .LR.cc.

Examples

morie_christoffersen_cc(c(rep(0, 40), 1, 1, 1, 0, 0, 1, rep(0, 54)), 0.05)$lr_ind
#> [1] 8.561074