getInitParams: Get multiple starting parameters for EM

In DOBAD: Analysis of Discretely Observed Linear Birth-and-Death(-and-Immigration) Markov Chains

Description

This is an ad-hoc function that has some hardwired rules for grabbing a few starting parameter values given one passed in one. You pass in your summary data and it makes a basic guess at starting parameters and then flips them and scales them in various ways depending on what numInitParams is. It returns between 1 and 6 different values (pairs) of starting parameters. Useful to get more than one parameter for if you are automating the EM in some way. Otherwise it just gives you the smart starting guess.

Usage

getInitParams(numInitParams=1, summary.PO, T, beta.immig, diffScale)

Arguments

numInitParams	How many parameters you want returned; between 1 and 6. Note: the paramaters after the 1st are fairly arbitrary.
summary.PO	Summary data from partially observed process. "Nplus", "Nminus", and "Holdtime" should be names in that order of number of observed jumsp up, jumps down, and Holding time.
T	total time of chain.
beta.immig	Scalar multiple of lambda that gives you the immigration rate, ie immigrate = beta.immig * birthrate.
diffScale	Note that we don't have a solution in the case mu == lambda. So if the two are close then numerical differentiation requires smaller values essentially. So usually pass something like "100*dr" where dr is the value that's passed through the add.joint.mean.*, etc (called delta) for numeric differentiation.

DOBAD documentation built on May 2, 2019, 3:04 a.m.