clusterMix {bayesm} | R Documentation |
clusterMix
uses MCMC draws of indicator variables from a normal component mixture model to cluster observations based on a similarity matrix.
clusterMix(zdraw, cutoff=0.9, SILENT=FALSE, nprint=BayesmConstant.nprint)
zdraw |
R x nobs array of draws of indicators |
cutoff |
cutoff probability for similarity (def: |
SILENT |
logical flag for silent operation (def: |
nprint |
print every nprint'th draw (def: |
Define a similarity matrix, Sim with Sim[i,j]=1
if observations i and j are in same component. Compute the posterior mean of Sim over indicator draws.
Clustering is achieved by two means:
Method A: Find the indicator draw whose similarity matrix minimizes loss(E[Sim]-Sim(z)), where loss is absolute deviation.
Method B: Define a Similarity matrix by setting any element of E[Sim] = 1 if E[Sim] > cutoff. Compute the clustering scheme associated with this "windsorized" Similarity matrix.
A list containing:
clustera: |
indicator function for clustering based on method A above |
clusterb: |
indicator function for clustering based on method B above |
This routine is a utility routine that does not check the input arguments for proper dimensions and type.
Peter Rossi, Anderson School, UCLA, perossichi@gmail.com.
For further discussion, see Bayesian Statistics and Marketing by Rossi, Allenby, and McCulloch Chapter 3.
http://www.perossi.org/home/bsm-1
if(nchar(Sys.getenv("LONG_TEST")) != 0) { ## simulate data from mixture of normals n = 500 pvec = c(.5,.5) mu1 = c(2,2) mu2 = c(-2,-2) Sigma1 = matrix(c(1,0.5,0.5,1), ncol=2) Sigma2 = matrix(c(1,0.5,0.5,1), ncol=2) comps = NULL comps[[1]] = list(mu1, backsolve(chol(Sigma1),diag(2))) comps[[2]] = list(mu2, backsolve(chol(Sigma2),diag(2))) dm = rmixture(n, pvec, comps) ## run MCMC on normal mixture Data = list(y=dm$x) ncomp = 2 Prior = list(ncomp=ncomp, a=c(rep(100,ncomp))) R = 2000 Mcmc = list(R=R, keep=1) out = rnmixGibbs(Data=Data, Prior=Prior, Mcmc=Mcmc) ## find clusters begin = 500 end = R outclusterMix = clusterMix(out$nmix$zdraw[begin:end,]) ## check on clustering versus "truth" ## note: there could be switched labels table(outclusterMix$clustera, dm$z) table(outclusterMix$clusterb, dm$z) }