opt.GRW.shift: Fit random walk model with shift(s) in generating parameters

Description

Fit random walk model with shift(s) in generating parameters

Usage

opt.GRW.shift(y, ng = 2, minb = 7, model = 1, pool = TRUE, silent = FALSE)

Value

a paleoTSfit object

Arguments

y: a paloeTS object
ng: number of segments in the sequence
minb: minimum number of populations in each segment
model: numeric, specifies exact evolutionary model; see Details
pool: if TRUE, sample variances are substituted with their pooled estimate
silent: logical, if TRUE, progress updates are suppressed

Details

Fits a model in which a sequence is divided into two or more segments and trait evolution proceeds as a general random walk, with each segment (potentially) getting its own generating parameters (mstep, vstep).

This function tests for shifts after each population, subject to the constraint that the number of populations in each segment is always >= minb. The shiftpoint yielding the highest log-likelihood is returned as the solution, along with the log-likelihoods (all.logl of all tested shift points (GG).

Different variants of the model can be specified by the model argument:

model = 1: mstep is separate across segments; vstep is shared
model = 2: mstep is shared across segments; vstep is separate
model = 3: mstep is set to zero (unbiased random walk); vstep is separate across segments
model = 4: mstep and vstep are both separate across segments

Examples

Run this code

x <- sim.GRW.shift(ns = c(15,15), ms = c(0, 1), vs = c(0.1,0.1))
w.sep <- opt.GRW.shift(x, ng = 2, model = 4)
w.sameVs <- opt.GRW.shift(x, ng = 2, model = 1)
compareModels(w.sep, w.sameVs)
plot(x)
abline(v = x$tt[16], lwd = 3)  # actual shift point
abline(v = x$tt[w.sameVs$par["shift1"]], lty = 3, col = "red", lwd = 2) # inferred shift point