Publisher review:Estimation, Simulation and Forecasting of an Autoregressive Markov Switching Model in Matlab - Functions to estimate, simulate and forecast a MS(k)-AR(p) model in Matlab This submission provides functions for estimation, simulation and forecasting of Autoregressive Markov Switching models with k states and p lags in Matlab. The model coded switches in AR parameters, mean constant and also in the model's standard deviation.The idea behind Markov switching models is to see the changes of states in the data as stochastic events and not deterministic ones. This class of models can give important information of how the states of the data (if found any) are switching and what's the duration of each.
The file comes with example data and also example scripts for simulation, estimation (or both combined) and forecasting. It is also a very nice example of how to setup and estimate econometric models in Matlab as a constrained optimization problem.From what I've played with it, fmincon() is doing a pretty good job at estimating the models. You can check this out by running the script Example_Script_Simul_and_Fit_2_States.mFell free to send any comments/suggestions.Perlin[Spec_Output]=MS_AR_Fit(x,ar,k)Input:x - Time series (vector)ar - Maximum lag of autoregression equationk - Number of statesOutput:Spec_Output - Output Structure with coefficients, probabilities over time, log likelihood of model among other things[Simul_Out]=MS_AR_Sim(nr,Coeff,k)Input: nr - Number of rows (number of time periods)Coeff - Coeff structure with all the coefficients k - number of StatesIn the coeff struct, each collum represent each state. Take a look at the example_Script_Simul.m, just follow it and you'll find very easy to simulate your own flavor (value of coefficients)Output: Simul_Out - A structure with following fields:Sim_x - Simulated time seriesCoeff - Structure with all coefficientsStates- The "true" simulated states through time[meanFor,stdFor]=MS_AR_For(Spec_Output,x)Input: Spec_Output - Specification output from estimation (check MS_AR_Fit.m)x - Time series to be forecastedOutput: meanFor - mean forecast at t 1stdFor - sigma forecast at t 1
Estimation, Simulation and Forecasting is a Matlab script for Earth Sciences scripts design by Marcelo Scherer Perlin.
It runs on following operating system: Windows / Linux / Mac OS / BSD / Solaris.
Operating system:Windows / Linux / Mac OS / BSD / Solaris
Estimation, Simulation and Forecasting script details 
