The correlated Wishart model provides the standard benchmark when analyzing time series. Unfortunately, the real case which is the most relevant one in applications, poses serious challenges for analytical calculations. Often these challenges are due to square root singularities which cannot be handled using common random matrix techniques. We present a new way to tackle this issue. For large but finite matrix dimensions, we show that statistical quantities in the bulk of the real correlated Wishart model with arbitrary empirical eigenvalues approach those of a real correlated Wishart model with matrix dimensions twice as large and doubly degenerate empirical eigenvalues. Our observation is based on an analytical study using supersymmetry and is confirmed by numerical simulations.
View on arXiv