phyclust.logL: Log-Likelihood of phyclust

In snoweye/phyclust: Phylogenetic Clustering (Phyloclustering)

Log-Likelihood of phyclust

Description

This computes a log-likelihood value of phyclust.

Usage

phyclust.logL(X, ret.phyclust = NULL, K = NULL, Eta = NULL,
    Mu = NULL, pi = NULL, kappa = NULL, Tt = NULL,
    substitution.model = NULL, identifier = NULL, code.type = NULL,
    label = NULL)

Arguments

X	nid/sid matrix with N rows/sequences and L columns/sites.
ret.phyclust	an object with the class phyclust.
K	number of clusters.
Eta	proportion of subpopulations, \eta_k, length = K, sum to 1.
Mu	centers of subpopulations, dim = K\times L, each row is a center.
pi	equilibrium probabilities, each row sums to 1.
kappa	transition and transversion bias.
Tt	total evolution time, t.
substitution.model	substitution model.
identifier	identifier.
code.type	code type.
label	label of sequences for semi-supervised clustering.

Details

X should be a numerical matrix containing sequence data that can be transfered by code2nid or code2sid.

Either input ret.phyclust or all other arguments for this function. ret.phyclust can be obtain either from an EM iteration of phyclust or from a M step of phyclust.m.step.

If label is inputted, the label information will be used to calculate log likelihood (complete-data), even the ret.phyclust is the result of unsupervised clustering.

Value

This function returns a log-likelihood value of phyclust.

Author(s)

Wei-Chen Chen [e-mail]

References

Phylogenetic Clustering Website: https://snoweye.github.io/phyclust/

See Also

phyclust, phyclust.em.step.

Examples

## Not run: 
library(phyclust, quiet = TRUE)

EMC.1 <- .EMC
EMC.1$EM.iter <- 1
# the same as EMC.1 <- .EMControl(EM.iter = 1)
X <- seq.data.toy$org

ret.1 <- phyclust(X, 2, EMC = EMC.1)
phyclust.logL(X, ret.phyclust = ret.1)

# For semi-supervised clustering.
semi.label <- rep(0, nrow(X))
semi.label[1:3] <- 1
phyclust.logL(X, ret.phyclust = ret.1, label = semi.label)

## End(Not run)

snoweye/phyclust documentation built on Sept. 12, 2023, 5 a.m.