merge_rotations {PhylogeneticEM}

Merge fits from independent runs of PhyloEM.

Description

merge_rotations takes several fits from PhyloEM, and merge them according to the best score (maximum likelihood or least squares). For each number of shifts, The datasets needs to be equal up to a rotation. This is tested thanks to a QR decomposition, see function find_rotation.

Usage

merge_rotations(..., method.selection = NULL, tol = NULL)

Arguments

...

objects of class PhyloEM fitted on datasets that are equal up to a rotation.

method.selection

(optional) selection method to be applied to the merged fit. See params_process.PhyloEM.

tol

(optional) relative numerical tolerance. See find_rotation.

Value

An object of class PhyloEM, result of the merge.

Examples

## Not run: 
## Load Data
data(monkeys)
## Run method
# Note: use more alpha values for better results.
res <- PhyloEM(Y_data = monkeys$dat,        ## data
               phylo = monkeys$phy,         ## phylogeny
               process = "scOU",            ## scalar OU
               random.root = TRUE,          ## root is stationary
               stationary.root = TRUE,
               K_max = 10,                  ## maximal number of shifts
               nbr_alpha = 4,               ## number of alpha values
               parallel_alpha = TRUE,       ## parallelize on alpha values
               Ncores = 2)
## Rotate dataset
rot <- matrix(c(cos(pi/4), -sin(pi/4), sin(pi/4), cos(pi/4)), nrow= 2, ncol = 2)
Yrot <- t(rot) %*% monkeys$dat
rownames(Yrot) <- rownames(monkeys$dat)
## Fit rotated dataset
# Note: use more alpha values for better results.
res_rot <- PhyloEM(Y_data = Yrot,               ## rotated data
                   phylo = monkeys$phy,         
                   process = "scOU",            
                   random.root = TRUE,          
                   stationary.root = TRUE,
                   K_max = 10,                  
                   nbr_alpha = 4,               
                   parallel_alpha = TRUE,       
                   Ncores = 2)
## Merge the two
res_merge <- merge_rotations(res, res_rot)
## Plot the selected result
plot(res_merge)
## Plot the model selection criterion
plot_criterion(res_merge)

## End(Not run)