Hi, I am looking for a package to estimate regime switching models (states following a markov chain). I found packages for Hidden Markov Models but I am looking for something a little different: In the HMM the conditional distribution of the observations (give the state) is a known distribution (normal or others), while the package I need should allow to set a conditional distribution (given the state) which can be still modelled (for example with mean-reversion or jump diffusion...). I think the theory under this estimation technique is in "James D. Hamilton, A New Approach to the Economic Analysis of Nonstationary Time Series and the Business Cycle (1989)" Thanks very much for any help! Valentina Bonetti Master student at Bocconi University, Milan [[alternative HTML version deleted]]
Hello, On 10/30/07, valentina bonetti <valentina.bonetti at gmail.com> wrote:> Hi, > > I am looking for a package to estimate regime switching models (states > following a markov chain). > I found packages for Hidden Markov Models but I am looking for something a > little different: In the HMM the conditional distribution of the > observations (give the state) is a known distribution (normal or others), > while the package I need should allow to set a conditional distribution > (given the state) which can be still modelled (for example with > mean-reversion or jump diffusion...). > > I think the theory under this estimation technique is in "James D. Hamilton, > A New Approach to the Economic Analysis of Nonstationary Time Series and the > Business Cycle (1989)"Not sure if this is the answer to your question, but check these references to "markov" in the Task Views [1]. [1] http://www.google.com/search?client=opera&rls=en&q=markov+site:http://cran.r-project.org/src/contrib/Views/&sourceid=opera&ie=utf-8&oe=utf-8 Regards, Liviu
Dear Sir, I saw a question from a member in 2007 asking for regime switching model and I am looking for the same kind of model. I also saw a "project" named RSNL but I don't know if this one is available and if we can have access to the code to change things if needed... actually I used to build my own models and not ask somebody but this one is relatively easy but long to implement so I wanted to skip this phase :-) I thus wanted to know if this model was available for download ? Thanks for your answer Mary