segmixt {segclust} | R Documentation |
segmentation of a signal when considering changes in a mixture of Gaussian vectors
out <- segmixt(x,P,Kmax,phi,lmin=1,lmax=length(x))
x |
data vector |
P |
number of clusters |
Kmax |
maximum number of segments |
phi |
parameters of the mixture |
lmin |
minimal segment length, default value lmin = 1 |
lmax |
maximal segment length, default value lmax = length(x) |
This function can be used for a segmentation/clustering model with P clusters and up to Kmax segments. phi gives the parameters of the mixture (P means, P standard deviations, P mixture proportions). It uses dynamic programming to find the best breakpoints, and is based on the calculus of the incomplete-data log-likelihood:
J.est_{K,P} = sum_{k=1}^{K} log [ sum_{p=1}^{P} (pi_p f(y^k;theta_p)) ]
where f(y^k;theta_p) is the density of a Gaussian vector of length n_k.
J.est |
incomplete-data log-likelihood for a segmentation/clustering model with P clusters and up to Kmax segments |
t.est |
estimated positions of breakpoints for a segmentation/clustering model up to Kmax segments |
F. Picard, M. Hoebecke
Picard, F., Robin, S., Lebarbier, E., & Daudin, J. -J. (2007). A segmentation/clustering model for the analysis of array CGH data. Biometrics, 63(3) 758-766