cbird {cbird} | R Documentation |
This function conducts the clustering of binary data with reducing the dimensionality (CLUSBIRD) proposed by Yamamoto and Hayashi (2015).
cbird(Y, N.comp, N.clust, lambda=0, N.ite=10000, N.random=1, show.random.ite=FALSE, eps=0.0001, mc.cores=1)
Y |
Binary data matrix (N * D), where N denotes sample size and D denotes the number of binary variables (0 or 1). |
N.comp |
The number of component to be extracted. |
N.clust |
The number of mixture components, which corresponds to the number of clusters. |
lambda |
A tuning parameter of an L1 penalty for loadings. A non-negative real value should be used as the value of lambda. |
N.ite |
The number of maximum of iterations for the EM algorithm. |
N.random |
The number of random sets of parameters for initial random starts. |
show.random.ite |
If |
eps |
The criterion for the convergence of the alternating least-squares
algorithm, which should be specified as a positive real value. If
the difference between the values of penalized log likelihood
functions of successive iteration is smaller than |
mc.cores |
If |
F |
An estimated component score matrix for cluster centroids. |
A |
An estimated loading matrix. |
mu |
Estimated mean values in the subspace. |
U |
The cluster assignment matrix (N * |
g |
The estimated mixture probability. |
n.ite |
The number of iteration needed for convergence. |
loss |
The value of log likelihood with L1 penalty. |
bic |
The value of BIC. |
LL |
The value of log likelihood. |
cluster |
Estimated clusters where subjects were assigned to |
ptime |
Time for calculation |
Michio Yamamoto
michio.koko@gmail.com
Yamamoto, M. and Hayashi, K. (2015). Clustering of multivariate binary data with dimension reduction via L1-regularized maximization. Pattern Recognition, 48, 3959-3968.
##Random Binary Data (unmeaningful example) ##100 subjects and 20 variables ##Consider three mixture components in the data set.seed(1) Y <- matrix(rbinom(100 * 20, 1, 0.5), 100, 20) out <- cbird(Y, 2, 3) est <- EstScore(Y, out$A, out$mu)