Latent Model Extreme Value Index Estimation
We propose a novel strategy for multivariate extreme value index estimation. In applications such as finance, volatility and risk present in the components of a multivariate time series are often driven by the same underlying factors, such as the subprime crisis in the US. To estimate the latent risk, we apply a two-stage procedure. First, a set of independent latent series is estimated using a method of latent variable analysis. Then, univariate risk measures are estimated individually for the latent series to assess their contribution to the overall risk. As our main theoretical contribution, we derive conditions under which the effect of the first step to the asymptotic behavior of the risk estimators is negligible. Simulations demonstrate the theory under both i.i.d. and dependent data, and an application into financial data illustrates the usefulness of the method in extracting joint sources of risk in practice.