optimParams: optimParams

In morpheus: Estimate Parameters of Mixtures of Logistic Regressions

optimParams

Description

Wrapper function for OptimParams class

Usage

optimParams(X, Y, K, link = c("logit", "probit"), M = NULL, nc = 0)

Arguments

X	Data matrix of covariables
Y	Output as a binary vector
K	Number of populations.
link	The link type, 'logit' or 'probit'.
M	the empirical cross-moments between X and Y (optional)
nc	Number of cores (default: 0 to use all)

Value

An object 'op' of class OptimParams, initialized so that op$run(theta0) outputs the list of optimized parameters

• p: proportions, size K

• beta: regression matrix, size dxK

• b: intercepts, size K

theta0 is a list containing the initial parameters. Only beta is required (p would be set to (1/K,...,1/K) and b to (0,...0)).

See Also

multiRun to estimate statistics based on beta, and generateSampleIO for I/O random generation.

Examples

# Optimize parameters from estimated mu
io <- generateSampleIO(100,
  1/2, matrix(c(1,-2,3,1),ncol=2), c(0,0), "logit")
mu <- computeMu(io$X, io$Y, list(K=2))
o <- optimParams(io$X, io$Y, 2, "logit")
## Not run: 
theta0 <- list(p=1/2, beta=mu, b=c(0,0))
par0 <- o$run(theta0)
# Compare with another starting point
theta1 <- list(p=1/2, beta=2*mu, b=c(0,0))
par1 <- o$run(theta1)
# Look at the function values at par0 and par1:
o$f( o$linArgs(par0) )
o$f( o$linArgs(par1) )
## End(Not run)

morpheus documentation built on Feb. 16, 2023, 10:01 p.m.