This function performs a gridsearch to find a good starting value for the EM algorithm. A good starting value for the EM algorithm is one for which all observations have strictly positive density (the higher the better)
GridSearchS0(family, y, params, size = 0, lbpdf = 0)accepted parameter set
distribution name; run the function distributions() for help
observations
list of six vectors named (p1, p2, p3, p4, p5, p6). Each corresponding to a parameter of the distribution (additionnal parameters will be ignored). For example : params = list(p1=c(0.5, 5, 0.5), p2=c(1, 5, 1), p3=c(0.1, 0.9, 0.1), p4=c(1,1,1), p5=c(1,1,1), p6=c(1,1,1)) where p1 is the grid of value for the first parameter.
additional parameter for some discrete distributions; run the command distributions() for help
minimal acceptable value of the density; (should be >= 0)
family = "gaussian"
Q = matrix(c(0.8, 0.3, 0.2, 0.7), 2, 2) ;
theta = matrix(c(-1.5, 1.7, 1, 1),2,2) ;
sim = SimHMMGen(theta, size=0, Q, ZI=0,"gaussian", 50)$SimData ;
params = list(p1=c(-2, 2, 0.5), p2=c(1, 5, 1), p3=c(1, 1, 1), p4=c(1,1,1), p5=c(1,1,1), p6=c(1,1,1))
accepted_params = GridSearchS0(family, sim, params)
Run the code above in your browser using DataLab