Next: Bias due to portfolio
Up: Problems of pathological VaR
Previous: Near singularity
Even if we have very good models (and even if n>d), simply due to
inevitable stochastic errors in the estimation of 
it is possible
for the covariance matrix to become near singular and admit portfolios
for which risk is pathologically underestimated.
More often however, stochastic errors will lead to spurious minima in the VaR
- which can also lead to a systematic underestimation of risk as explained
in the next subsection.