lmap (version 0.2.4)

lpca: Logistic (Restricted) PCA

Description

This function runs: logistic principal component analysis (if X = NULL) logistic reduced rank regression (if X != NULL)

Usage

lpca(
  Y,
  X = NULL,
  S = 2,
  start = NULL,
  dim.indic = NULL,
  eq = FALSE,
  lambda = FALSE,
  maxiter = 65536,
  dcrit = 1e-06
)

Value

This function returns an object of the class lpca with components:

call

Call to the function

Y

Matrix Y from input

Xoriginal

Matrix X from input

X

Scaled X matrix

mx

Mean values of X

sdx

Standard deviations of X

ynames

Variable names of responses

xnames

Variable names of predictors

probabilities

Estimated values of Y

m

main effects

U

matrix with coordinates for row-objects

B

matrix with regression weight (U = XB)

V

matrix with vectors for items/responses

iter

number of main iterations from the MM algorithm

deviance

value of the deviance at convergence

npar

number of estimated parameters

AIC

Akaike's Information Criterion

BIC

Bayesian Information Criterion

Arguments

Y

An N times R binary matrix .

X

An N by P matrix with predictor variables

S

Positive number indicating the dimensionality of the solution

start

Option to provide starting values (list with m, U or B, and V)

dim.indic

An R by S matrix indicating which response variable pertains to which dimension

eq

Only applicable when dim.indic not NULL; equality restriction on regression weighhts per dimension

lambda

if TRUE does lambda scaling (see Understanding Biplots, p24)

maxiter

maximum number of iterations

dcrit

convergence criterion

Examples

if (FALSE) {
data(dataExample_lpca)
Y = as.matrix(dataExample_lpca[, 1:8])
X = as.matrix(dataExample_lpca[, 9:13])
# unsupervised
output = lpca(Y = Y, S = 2)
}