October 8th, 2009, 4:05 pm
QuoteOriginally posted by: willsmithMarkov chains / regimes are often useful in volatility modelling, volatility sits happily in a quiet regime and boom, suddenly it switches to a noisy regime. Then after a while, back again. For example, electricity spot price modelling often uses this.I guess it's useful anywhere with regimes which have different parameters. But much harder to calibrate than a single regime. Instead of constant parameters, even in the case of 2 regimes, you have to calibrate (1) parameters in regime A (2) parameters in regime B (3) prob(move from regime A to regime B in a given time period) (4) prob(move from regime B to regime A in a given time period).I'd say, don't use markov chains / regimes unless you see you need them. They are not a solution looking for a problem!also:(5) prob of currently being in regime A( or B)