SpaTimeClus (version 1.0.1)

spatimeclus: This function performs the maximum likelihood estimation for a known model in clustering

Description

This function performs the maximum likelihood estimation for a known model in clustering

Usage

spatimeclus(obs, G, K, Q, map = NULL, m = 1:(dim(obs)[3]), crit = "BIC",
  tol = 0.001, param = NULL, nbcores = 1, nbinitSmall = 500,
  nbinitKept = 50, nbiterSmall = 20, nbiterKept = 500)

Arguments

obs

array It contains the observations to cluster where the dimensions are respectively: number of the observation, site of the observation, time of the observation.

G

numeric. It defines possible numbers of components.

K

numeric. It defines possible numbers of regressions per components

Q

numeric. It defines possible degrees of regressions.

map

matrix. It gives the spatial coordiantes of each site.

m

numeric. It indicates the moments of observations (optional, default is 1:T).

crit

character. It indicates the criterion used for the model selection ("AIC", "BIC" or "ICL", optional, default is "BIC").

tol

numeric. The algorithm is stopped when the loglikelihood increases less than tol during two successive iterations (optional, default is 0.001).

param

list of '>STCparam. It gives the initial values of the EM algorithm (optional, starting point are sampled at random).

nbcores

numeric. It defines the numerber of cores used by the alogrithm, only for Linux and Mac (optional, default is 1).

nbinitSmall

numeric. It defines the number of random initializations (optional, default is 500).

nbinitKept

numeric. It defines the number of chains estimated until convergence (optional, default is 50).

nbiterSmall

numeric. It defines the number of iterations before keeping the nbinitKept best chains (optional, default is 20).

nbiterKept

numeric. It defines the maximum number of iterations before to stop the algorith; (optional, default is 500).

Value

Returns an instance of '>STCresults.

Examples

# NOT RUN {
data(airparif)

# Clustering of the data by considering the spatial dependencies
res.spa <- spatimeclus(airparif$obs,  G=3, K=4, Q=4, map = airparif$map,
 nbinitSmall=50, nbinitKept=5, nbiterSmall=5)
summary(res.spa)

# Clustering of the data without considering the spatial dependencies
res.nospa <- spatimeclus(airparif$obs,  G=3, K=4, Q=4, nbinitSmall=50, nbinitKept=5, nbiterSmall=5)
summary(res.nospa)
# }