Extreme events: blending PCA and ICA [10]

Go to: Summary | Previous | Next   
Bullet points include: E.g. consider two return series in tandem, bucket each into quantile boxes and plot the number of times each quantile box pairing occurs. Maybe include all possible unit +/- stances to make four corners of the plot symmetrical? Aggregate plots for all possible sector pairs? E.g. chart opposite based on monthly (log) sector relative price movements for 23 MSCI AC-World sectors with complete series between (30/05/96 and 28/02/09). Strong evidence that correlations “tend to unity” in stressed times? Or merely that we are mixing different distributions together?

Contents | Prev | Next | Library

Desktop view | Switch to Mobile